In preparation for BPF-to-BPF calls in offloaded programs, add a new
function attribute to the struct bpf_prog_offload_ops so that drivers
supporting eBPF offload can hook at the end of program verification, and
potentially extract information collected by the verifier.
Implement a minimal callback (returning 0) in the drivers providing the
structs, namely netdevsim and nfp.
This will be useful in the nfp driver, in later commits, to extract the
number of subprograms as well as the stack depth for those subprograms.
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jiong Wang <jiong.wang@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Allow programs and maps to be re-used across different netdevs,
as long as they belong to the same struct bpf_offload_dev.
Update the bpf_offload_prog_map_match() helper for the verifier
and export a new helper for the drivers to use when checking
programs at attachment time.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Create a higher-level entity to represent a device/ASIC to allow
programs and maps to be shared between device ports. The extra
work is required to make sure we don't destroy BPF objects as
soon as the netdev for which they were loaded gets destroyed,
as other ports may still be using them. When netdev goes away
all of its BPF objects will be moved to other netdevs of the
device, and only destroyed when last netdev is unregistered.
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>
Currently we have two lists of offloaded objects - programs and maps.
Netdevice unregister notifier scans those lists to orphan objects
associated with device being unregistered. This puts unnecessary
(even if negligible) burden on all netdev unregister calls in BPF-
-enabled kernel. The lists of objects may potentially get long
making the linear scan even more problematic. There haven't been
complaints about this mechanisms so far, but it is suboptimal.
Instead of relying on notifiers, make the few BPF-capable drivers
register explicitly for BPF offloads. The programs and maps will
now be collected per-device not on a global list, and only scanned
for removal when driver unregisters from BPF offloads.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
A set of new API functions exported for the drivers will soon use
'bpf_offload_dev_' as a prefix. Rename the bpf_offload_dev_match()
which is internal to the core (used by the verifier) to avoid any
confusion.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
BPF_MAP_TYPE_PERF_EVENT_ARRAY is special as far as offload goes.
The map only holds glue to perf ring, not actual data. Allow
non-offloaded perf event arrays to be used in offloaded programs.
Offload driver can extract the events from HW and put them in
the map for user space to retrieve.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: Jiong Wang <jiong.wang@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tell user space about device on which the map was created.
Unfortunate reality of user ABI makes sharing this code
with program offload difficult but the information is the
same.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The special handling of different map types is left to the driver.
Allow offload of array maps by simply adding it to accepted types.
For nfp we have to make sure array elements are not deleted.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
For host JIT, there are "jited_len"/"bpf_func" fields in struct bpf_prog
used by all host JIT targets to get jited image and it's length. While for
offload, targets are likely to have different offload mechanisms that these
info are kept in device private data fields.
Therefore, BPF_OBJ_GET_INFO_BY_FD syscall needs an unified way to get JIT
length and contents info for offload targets.
One way is to introduce new callback to parse device private data then fill
those fields in bpf_prog_info. This might be a little heavy, the other way
is to add generic fields which will be initialized by all offload targets.
This patch follow the second approach to introduce two new fields in
struct bpf_dev_offload and teach bpf_prog_get_info_by_fd about them to fill
correct jited_prog_len and jited_prog_insns in bpf_prog_info.
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Jiong Wang <jiong.wang@netronome.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Daniel suggests it would be more logical for bpf_offload_dev_match()
to return false is either the program or the map are not offloaded,
rather than treating the both not offloaded case as a "matching
CPU/host device".
This makes no functional difference today, since verifier only calls
bpf_offload_dev_match() when one of the objects is offloaded.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
BPF map offload follow similar path to program offload. At creation
time users may specify ifindex of the device on which they want to
create the map. Map will be validated by the kernel's
.map_alloc_check callback and device driver will be called for the
actual allocation. Map will have an empty set of operations
associated with it (save for alloc and free callbacks). The real
device callbacks are kept in map->offload->dev_ops because they
have slightly different signatures. Map operations are called in
process context so the driver may communicate with HW freely,
msleep(), wait() etc.
Map alloc and free callbacks are muxed via existing .ndo_bpf, and
are always called with rtnl lock held. Maps and programs are
guaranteed to be destroyed before .ndo_uninit (i.e. before
unregister_netdev() returns). Map callbacks are invoked with
bpf_devs_lock *read* locked, drivers must take care of exclusive
locking if necessary.
All offload-specific branches are marked with unlikely() (through
bpf_map_is_dev_bound()), given that branch penalty will be
negligible compared to IO anyway, and we don't want to penalize
SW path unnecessarily.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Add a helper to check if netdev could be found and whether it
has .ndo_bpf callback. There is no need to check the callback
every time it's invoked, ndos can't reasonably be swapped for
a set without .ndp_bpf while program is loaded.
bpf_dev_offload_check() will also be used by map offload.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
With map offload coming, we need to call program offload structure
something less ambiguous. Pure rename, no functional changes.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Report to the user ifindex and namespace information of offloaded
programs. If device has disappeared return -ENODEV. Specify the
namespace using dev/inode combination.
CC: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Bound programs are quite useless after their device disappears.
They are simply waiting for reference count to go to zero,
don't list them in BPF_PROG_GET_NEXT_ID by freeing their ID
early.
Note that orphaned offload programs will return -ENODEV on
BPF_OBJ_GET_INFO_BY_FD so user will never see ID 0.
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>
All bpf offload operations should now be under bpf_devs_lock,
it's safe to free and clear the entire offload structure,
not only the netdev pointer.
__bpf_prog_offload_destroy() will no longer be called multiple
times.
Suggested-by: Alexei Starovoitov <ast@kernel.org>
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>
To allow verifier instruction callbacks without any extra locking
NETDEV_UNREGISTER notification would wait on a waitqueue for verifier
to finish. This design decision was made when rtnl lock was providing
all the locking. Use the read/write lock instead and remove the
workqueue.
Verifier will now call into the offload code, so dev_ops are moved
to offload structure. Since verifier calls are all under
bpf_prog_is_dev_bound() we no longer need static inline implementations
to please builds with CONFIG_NET=n.
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>
We don't need the RTNL lock for all operations on offload state.
We only need to hold it around ndo calls. The device offload
initialization doesn't require it. The soon-to-come querying
of the offload info will only need it partially. We will also
be able to remove the waitqueue in following patches.
Use struct rw_semaphore because map offload will require sleeping
with the semaphore held for read.
Suggested-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
I forgot to add a license on kernel/bpf/offload.c. Luckily I'm
still the only author so make it explicitly GPLv2.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This reverts commit bd601b6ada ("bpf: report offload info to user
space"). The ifindex by itself is not sufficient, we should provide
information on which network namespace this ifindex belongs to.
After considering some options we concluded that it's best to just
remove this API for now, and rework it in -next.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
We are currently destroying the device offload state when device
moves to another net namespace. This doesn't break with current
NFP code, because offload state is not used on program removal,
but it's not correct behaviour.
Ignore the device unregister notifications on namespace move.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
bpf_target_prog seems long and clunky, rename it to prog_ifindex.
We don't want to call this field just ifindex, because maps
may need a similar field in the future and bpf_attr members for
programs and maps are unnamed.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
We are currently only allowing attachment of device-bound
cls_bpf and XDP programs. Make this restriction explicit in
the BPF offload code. This way we can potentially reuse the
ifindex field in the future.
Since XDP and cls_bpf programs can only be loaded by admin,
we can drop the explicit capability check from offload code.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Offload state may get destroyed either because the device for which
it was constructed is going away, or because the refcount of bpf
program itself has reached 0. In both of those cases we will call
__bpf_prog_offload_destroy() to unlink the offload from the device.
We may in fact call it twice, which works just fine, but we should
make clear this is intended and caution others trying to extend the
function.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Extend struct bpf_prog_info to contain information about program
being bound to a device. Since the netdev may get destroyed while
program still exists we need a flag to indicate the program is
loaded for a device, even if the device is gone.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The fact that we don't know which device the program is going
to be used on is quite limiting in current eBPF infrastructure.
We have to reverse or limit the changes which kernel makes to
the loaded bytecode if we want it to be offloaded to a networking
device. We also have to invent new APIs for debugging and
troubleshooting support.
Make it possible to load programs for a specific netdev. This
helps us to bring the debug information closer to the core
eBPF infrastructure (e.g. we will be able to reuse the verifer
log in device JIT). It allows device JITs to perform translation
on the original bytecode.
__bpf_prog_get() when called to get a reference for an attachment
point will now refuse to give it if program has a device assigned.
Following patches will add a version of that function which passes
the expected netdev in. @type argument in __bpf_prog_get() is
renamed to attach_type to make it clearer that it's only set on
attachment.
All calls to ndo_bpf are protected by rtnl, only verifier callbacks
are not. We need a wait queue to make sure netdev doesn't get
destroyed while verifier is still running and calling its driver.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Quentin Monnet
Jakub Kicinski
Jiong Wang