Björn Töpel [Mon, 20 Jan 2020 09:29:17 +0000 (10:29 +0100)]
xsk, net: Make sock_def_readable() have external linkage
XDP sockets use the default implementation of struct sock's
sk_data_ready callback, which is sock_def_readable(). This function
is called in the XDP socket fast-path, and involves a retpoline. By
letting sock_def_readable() have external linkage, and being called
directly, the retpoline can be avoided.
Al Viro [Mon, 20 Jan 2020 23:28:58 +0000 (23:28 +0000)]
bpf: don't bother with getname/kern_path - use user_path_at
kernel/bpf/inode.c misuses kern_path...() - it's much simpler (and
more efficient, on top of that) to use user_path...() counterparts
rather than bothering with doing getname() manually.
====================
We are currently being somewhat inconsistent with the libbpf include paths,
which makes it difficult to move files from the kernel into an external
libbpf-using project without adjusting include paths.
Having the bpf/ subdir of $INCLUDEDIR in the include path has never been a
requirement for building against libbpf before, and indeed the libbpf pkg-config
file doesn't include it. So let's make all libbpf includes across the kernel
tree use the bpf/ prefix in their includes. Since bpftool skeleton generation
emits code with a libbpf include, this also ensures that those can be used in
existing external projects using the regular pkg-config include path.
This turns out to be a somewhat invasive change in the number of files touched;
however, the actual changes to files are fairly trivial (most of them are simply
made with 'sed'). The series is split to make the change for one tool subdir at
a time, while trying not to break the build along the way. It is structured like
this:
- Patch 1-3: Trivial fixes to Makefiles for issues I discovered while changing
the include paths.
- Patch 4-8: Change the include directives to use the bpf/ prefix, and updates
Makefiles to make sure tools/lib/ is part of the include path, but without
removing tools/lib/bpf
- Patch 9-11: Remove tools/lib/bpf from include paths to make sure we don't
inadvertently re-introduce includes without the bpf/ prefix.
Changelog:
v5:
- Combine the libbpf build rules in selftests Makefile (using Andrii's
suggestion for a make rule).
- Re-use self-tests libbpf build for runqslower (new patch 10)
- Formatting fixes
v4:
- Move runqslower error on missing BTF into make rule
- Make sure we don't always force a rebuild selftests
- Rebase on latest bpf-next (dropping patch 11)
v3:
- Don't add the kernel build dir to the runqslower Makefile, pass it in from
selftests instead.
- Use libbpf's 'make install_headers' in selftests instead of trying to
generate bpf_helper_defs.h in-place (to also work on read-only filesystems).
- Use a scratch builddir for both libbpf and bpftool when building in selftests.
- Revert bpf_helpers.h to quoted include instead of angled include with a bpf/
prefix.
- Fix a few style nits from Andrii
v2:
- Do a full cleanup of libbpf includes instead of just changing the
bpf_helper_defs.h include.
====================
selftests: Refactor build to remove tools/lib/bpf from include path
To make sure no new files are introduced that doesn't include the bpf/
prefix in its #include, remove tools/lib/bpf from the include path
entirely.
Instead, we introduce a new header files directory under the scratch tools/
dir, and add a rule to run the 'install_headers' rule from libbpf to have a
full set of consistent libbpf headers in $(OUTPUT)/tools/include/bpf, and
then use $(OUTPUT)/tools/include as the include path for selftests.
For consistency we also make sure we put all the scratch build files from
other bpftool and libbpf into tools/build/, so everything stays within
selftests/.
runsqslower: Support user-specified libbpf include and object paths
This adds support for specifying the libbpf include and object paths as
arguments to the runqslower Makefile, to support reusing the libbpf version
built as part of the selftests.
tools/runqslower: Remove tools/lib/bpf from include path
Since we are now consistently using the bpf/ prefix on #include directives,
we don't need to include tools/lib/bpf in the include path. Remove it to
make sure we don't inadvertently introduce new includes without the prefix.
samples/bpf: Use consistent include paths for libbpf
Fix all files in samples/bpf to include libbpf header files with the bpf/
prefix, to be consistent with external users of the library. Also ensure
that all includes of exported libbpf header files (those that are exported
on 'make install' of the library) use bracketed includes instead of quoted.
To make sure no new files are introduced that doesn't include the bpf/
prefix in its include, remove tools/lib/bpf from the include path entirely,
and use tools/lib instead.
Fixes: 72ed068845b7 ("selftests/bpf: Ensure bpf_helper_defs.h are taken from selftests dir") Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Jesper Dangaard Brouer <brouer@redhat.com> Acked-by: Andrii Nakryiko <andriin@fb.com> Link: https://lore.kernel.org/bpf/157952560911.1683545.8795966751309534150.stgit@toke.dk
Fix bpftool to include libbpf header files with the bpf/ prefix, to be
consistent with external users of the library. Also ensure that all
includes of exported libbpf header files (those that are exported on 'make
install' of the library) use bracketed includes instead of quoted.
To make sure no new files are introduced that doesn't include the bpf/
prefix in its include, remove tools/lib/bpf from the include path entirely,
and use tools/lib instead.
Fixes: 72ed068845b7 ("selftests/bpf: Ensure bpf_helper_defs.h are taken from selftests dir") Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Link: https://lore.kernel.org/bpf/157952560684.1683545.4765181397974997027.stgit@toke.dk
selftests: Use consistent include paths for libbpf
Fix all selftests to include libbpf header files with the bpf/ prefix, to
be consistent with external users of the library. Also ensure that all
includes of exported libbpf header files (those that are exported on 'make
install' of the library) use bracketed includes instead of quoted.
To not break the build, keep the old include path until everything has been
changed to the new one; a subsequent patch will remove that.
Fixes: 72ed068845b7 ("selftests/bpf: Ensure bpf_helper_defs.h are taken from selftests dir") Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Link: https://lore.kernel.org/bpf/157952560568.1683545.9649335788846513446.stgit@toke.dk
tools/runqslower: Use consistent include paths for libbpf
Fix the runqslower tool to include libbpf header files with the bpf/
prefix, to be consistent with external users of the library. Also ensure
that all includes of exported libbpf header files (those that are exported
on 'make install' of the library) use bracketed includes instead of quoted.
To not break the build, keep the old include path until everything has been
changed to the new one; a subsequent patch will remove that.
Fixes: 72ed068845b7 ("selftests/bpf: Ensure bpf_helper_defs.h are taken from selftests dir") Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Link: https://lore.kernel.org/bpf/157952560457.1683545.9913736511685743625.stgit@toke.dk
selftests: Pass VMLINUX_BTF to runqslower Makefile
Add a VMLINUX_BTF variable with the locally-built path when calling the
runqslower Makefile from selftests. This makes sure a simple 'make'
invocation in the selftests dir works even when there is no BTF information
for the running kernel. Do a wildcard expansion and include the same paths
for BTF for the running kernel as in the runqslower Makefile, to make it
possible to build selftests without having a vmlinux in the local tree.
Also fix the make invocation to use $(OUTPUT)/tools as the destination
directory instead of $(CURDIR)/tools.
tools/bpf/runqslower: Fix override option for VMLINUX_BTF
The runqslower tool refuses to build without a file to read vmlinux BTF
from. The build fails with an error message to override the location by
setting the VMLINUX_BTF variable if autodetection fails. However, the
Makefile doesn't actually work with that override - the error message is
still emitted.
Fix this by including the value of VMLINUX_BTF in the expansion, and only
emitting the error message if the *result* is empty. Also permit running
'make clean' even though no VMLINUX_BTF is set.
samples/bpf: Don't try to remove user's homedir on clean
The 'clean' rule in the samples/bpf Makefile tries to remove backup
files (ending in ~). However, if no such files exist, it will instead try
to remove the user's home directory. While the attempt is mostly harmless,
it does lead to a somewhat scary warning like this:
rm: cannot remove '~': Is a directory
Fix this by using find instead of shell expansion to locate any actual
backup files that need to be removed.
Fixes: 52031cf8b010 ("samples/bpf: allow make to be run from samples/bpf/ directory") Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Jesper Dangaard Brouer <brouer@redhat.com> Link: https://lore.kernel.org/bpf/157952560126.1683545.7273054725976032511.stgit@toke.dk
Hangbin Liu [Fri, 17 Jan 2020 10:06:56 +0000 (18:06 +0800)]
selftests/bpf: Skip perf hw events test if the setup disabled it
The same with commit 23941fde1605 ("selftests/bpf: skip nmi test when perf
hw events are disabled"), it would make more sense to skip the
test_stacktrace_build_id_nmi test if the setup (e.g. virtual machines) has
disabled hardware perf events.
Fixes: 044ba7e1b5a3 ("bpf: add selftest for stackmap with build_id in NMI context") Signed-off-by: Hangbin Liu <liuhangbin@gmail.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Link: https://lore.kernel.org/bpf/20200117100656.10359-1-liuhangbin@gmail.com
After commit 767abe73755f ("libbpf: Don't require root for
bpf_object__open()") we no longer load BTF during bpf_object__open(),
so let's remove the expectation from test_btf that the fd is not -1.
The test currently fails.
Brian Vazquez [Sun, 19 Jan 2020 19:40:40 +0000 (11:40 -0800)]
bpf: Fix memory leaks in generic update/delete batch ops
Generic update/delete batch ops functions were using __bpf_copy_key
without properly freeing the memory. Handle the memory allocation and
copy_from_user separately.
Fixes: 352625a1b08d ("bpf: Add generic support for update and delete batch ops") Reported-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Brian Vazquez <brianvv@google.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/bpf/20200119194040.128369-1-brianvv@google.com
KP Singh [Fri, 17 Jan 2020 21:28:25 +0000 (22:28 +0100)]
libbpf: Load btf_vmlinux only once per object.
As more programs (TRACING, STRUCT_OPS, and upcoming LSM) use vmlinux
BTF information, loading the BTF vmlinux information for every program
in an object is sub-optimal. The fix was originally proposed in:
The btf_vmlinux is populated in the object if any of the programs in
the object requires it just before the programs are loaded and freed
after the programs finish loading.
Andrii Nakryiko [Fri, 17 Jan 2020 06:07:59 +0000 (22:07 -0800)]
libbpf: Simplify BTF initialization logic
Current implementation of bpf_object's BTF initialization is very convoluted
and thus prone to errors. It doesn't have to be like that. This patch
simplifies it significantly.
This code also triggered static analysis issues over logically dead code due
to redundant error checks. This simplification should fix that as well.
YueHaibing [Thu, 16 Jan 2020 14:53:00 +0000 (22:53 +0800)]
bpf: Remove set but not used variable 'first_key'
kernel/bpf/syscall.c: In function generic_map_lookup_batch:
kernel/bpf/syscall.c:1339:7: warning: variable first_key set but not used [-Wunused-but-set-variable]
====================
Since commit 40db8320a5e2 ("xdp: Make devmap flush_list common for all map
instances"), devmap flushing is a global operation instead of tied to a
particular map. This means that with a bit of refactoring, we can finally fix
the performance delta between the bpf_redirect_map() and bpf_redirect() helper
functions, by introducing bulking for the latter as well.
This series makes this change by moving the data structure used for the bulking
into struct net_device itself, so we can access it even when there is not
devmap. Once this is done, moving the bpf_redirect() helper to use the bulking
mechanism becomes quite trivial, and brings bpf_redirect() up to the same as
bpf_redirect_map():
After this patch series, the only semantics different between the two variants
of the bpf() helper (apart from the absence of a map argument, obviously) is
that the _map() variant will return an error if passed an invalid map index,
whereas the bpf_redirect() helper will succeed, but drop packets on
xdp_do_redirect(). This is because the helper has no reference to the calling
netdev, so unfortunately we can't do the ifindex lookup directly in the helper.
Changelog:
v3:
- Switch two more fields to avoid a list_head spanning two cache lines
- Include Jesper's tracepoint patch
- Also rename xdp_do_flush_map()
- Fix a few nits from Maciej
v2:
- Consolidate code paths and tracepoints for map and non-map redirect variants
(Björn)
- Add performance data for 2-CPU test (Jesper)
- Move fields to avoid shifting cache lines in struct net_device (Eric)
====================
devmap: Adjust tracepoint for map-less queue flush
Now that we don't have a reference to a devmap when flushing the device
bulk queue, let's change the the devmap_xmit tracepoint to remote the
map_id and map_index fields entirely. Rearrange the fields so 'drops' and
'sent' stay in the same position in the tracepoint struct, to make it
possible for the xdp_monitor utility to read both the old and the new
format.
xdp: Use bulking for non-map XDP_REDIRECT and consolidate code paths
Since the bulk queue used by XDP_REDIRECT now lives in struct net_device,
we can re-use the bulking for the non-map version of the bpf_redirect()
helper. This is a simple matter of having xdp_do_redirect_slow() queue the
frame on the bulk queue instead of sending it out with __bpf_tx_xdp().
Unfortunately we can't make the bpf_redirect() helper return an error if
the ifindex doesn't exit (as bpf_redirect_map() does), because we don't
have a reference to the network namespace of the ingress device at the time
the helper is called. So we have to leave it as-is and keep the device
lookup in xdp_do_redirect_slow().
Since this leaves less reason to have the non-map redirect code in a
separate function, so we get rid of the xdp_do_redirect_slow() function
entirely. This does lose us the tracepoint disambiguation, but fortunately
the xdp_redirect and xdp_redirect_map tracepoints use the same tracepoint
entry structures. This means both can contain a map index, so we can just
amend the tracepoint definitions so we always emit the xdp_redirect(_err)
tracepoints, but with the map ID only populated if a map is present. This
means we retire the xdp_redirect_map(_err) tracepoints entirely, but keep
the definitions around in case someone is still listening for them.
With this change, the performance of the xdp_redirect sample program goes
from 5Mpps to 8.4Mpps (a 68% increase).
Since the flush functions are no longer map-specific, rename the flush()
functions to drop _map from their names. One of the renamed functions is
the xdp_do_flush_map() callback used in all the xdp-enabled drivers. To
keep from having to update all drivers, use a #define to keep the old name
working, and only update the virtual drivers in this patch.
xdp: Move devmap bulk queue into struct net_device
Commit 40db8320a5e2 ("xdp: Make devmap flush_list common for all map
instances"), changed devmap flushing to be a global operation instead of a
per-map operation. However, the queue structure used for bulking was still
allocated as part of the containing map.
This patch moves the devmap bulk queue into struct net_device. The
motivation for this is reusing it for the non-map variant of XDP_REDIRECT,
which will be changed in a subsequent commit. To avoid other fields of
struct net_device moving to different cache lines, we also move a couple of
other members around.
We defer the actual allocation of the bulk queue structure until the
NETDEV_REGISTER notification devmap.c. This makes it possible to check for
ndo_xdp_xmit support before allocating the structure, which is not possible
at the time struct net_device is allocated. However, we keep the freeing in
free_netdev() to avoid adding another RCU callback on NETDEV_UNREGISTER.
Because of this change, we lose the reference back to the map that
originated the redirect, so change the tracepoint to always return 0 as the
map ID and index. Otherwise no functional change is intended with this
patch.
After this patch, the relevant part of struct net_device looks like this,
according to pahole:
Revert bpf_helpers.h's change to include auto-generated bpf_helper_defs.h
through <> instead of "", which causes it to be searched in include path. This
can break existing applications that don't have their include path pointing
directly to where libbpf installs its headers.
There is ongoing work to make all (not just bpf_helper_defs.h) includes more
consistent across libbpf and its consumers, but this unbreaks user code as is
right now without any regressions. Selftests still behave sub-optimally
(taking bpf_helper_defs.h from libbpf's source directory, if it's present
there), which will be fixed in subsequent patches.
Fixes: 72ed068845b7 ("selftests/bpf: Ensure bpf_helper_defs.h are taken from selftests dir") Reported-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: Andrii Nakryiko <andriin@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20200117004103.148068-1-andriin@fb.com
Yonghong Song [Thu, 16 Jan 2020 17:40:04 +0000 (09:40 -0800)]
selftests/bpf: Fix test_progs send_signal flakiness with nmi mode
Alexei observed that test_progs send_signal may fail if run
with command line "./test_progs" and the tests will pass
if just run "./test_progs -n 40".
I observed similar issue with nmi subtest failure
and added a delay 100 us in Commit 5ecf703708fe
("tools/bpf: Add self tests for bpf_send_signal_thread()")
and the problem is gone for me. But the issue still exists
in Alexei's testing environment.
The current code uses sample_freq = 50 (50 events/second), which
may not be enough. But if the sample_freq value is larger than
sysctl kernel/perf_event_max_sample_rate, the perf_event_open
syscall will fail.
This patch changed nmi perf testing to use sample_period = 1,
which means trying to sampling every event. This seems fixing
the issue.
Andrii Nakryiko [Thu, 16 Jan 2020 00:55:49 +0000 (16:55 -0800)]
selftests/bpf: Add whitelist/blacklist of test names to test_progs
Add ability to specify a list of test name substrings for selecting which
tests to run. So now -t is accepting a comma-separated list of strings,
similarly to how -n accepts a comma-separated list of test numbers.
Additionally, add ability to blacklist tests by name. Blacklist takes
precedence over whitelist. Blacklisting is important for cases where it's
known that some tests can't pass (e.g., due to perf hardware events that are
not available within VM). This is going to be used for libbpf testing in
Travis CI in its Github repo.
Example runs with just whitelist and whitelist + blacklist:
====================
When a map is storing a kernel's struct, its
map_info->btf_vmlinux_value_type_id is set. The first map type
supporting it is BPF_MAP_TYPE_STRUCT_OPS.
This series adds support to dump this kind of map with BTF.
The first two patches are bug fixes which are only applicable to
bpf-next.
Please see individual patches for details.
v3:
- Remove unnecessary #include "libbpf_internal.h" from patch 5
v2:
- Expose bpf_find_kernel_btf() as a LIBBPF_API in patch 3 (Andrii)
- Cache btf_vmlinux in bpftool/map.c (Andrii)
====================
Martin KaFai Lau [Wed, 15 Jan 2020 23:00:44 +0000 (15:00 -0800)]
bpftool: Support dumping a map with btf_vmlinux_value_type_id
This patch makes bpftool support dumping a map's value properly
when the map's value type is a type of the running kernel's btf.
(i.e. map_info.btf_vmlinux_value_type_id is set instead of
map_info.btf_value_type_id). The first usecase is for the
BPF_MAP_TYPE_STRUCT_OPS.
Martin KaFai Lau [Wed, 15 Jan 2020 23:00:31 +0000 (15:00 -0800)]
libbpf: Expose bpf_find_kernel_btf as a LIBBPF_API
This patch exposes bpf_find_kernel_btf() as a LIBBPF_API.
It will be used in 'bpftool map dump' in a following patch
to dump a map with btf_vmlinux_value_type_id set.
bpf_find_kernel_btf() is renamed to libbpf_find_kernel_btf()
and moved to btf.c. As <linux/kernel.h> is included,
some of the max/min type casting needs to be fixed.
Martin KaFai Lau [Wed, 15 Jan 2020 23:00:25 +0000 (15:00 -0800)]
bpftool: Fix missing BTF output for json during map dump
The btf availability check is only done for plain text output.
It causes the whole BTF output went missing when json_output
is used.
This patch simplifies the logic a little by avoiding passing "int btf" to
map_dump().
For plain text output, the btf_wtr is only created when the map has
BTF (i.e. info->btf_id != 0). The nullness of "json_writer_t *wtr"
in map_dump() alone can decide if dumping BTF output is needed.
As long as wtr is not NULL, map_dump() will print out the BTF-described
data whenever a map has BTF available (i.e. info->btf_id != 0)
regardless of json or plain-text output.
In do_dump(), the "int btf" is also renamed to "int do_plain_btf".
Fixes: f371215a1ed1 ("bpftool: Match maps by name") Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Cc: Paul Chaignon <paul.chaignon@orange.com> Link: https://lore.kernel.org/bpf/20200115230025.1101828-1-kafai@fb.com
Martin KaFai Lau [Wed, 15 Jan 2020 23:00:19 +0000 (15:00 -0800)]
bpftool: Fix a leak of btf object
When testing a map has btf or not, maps_have_btf() tests it by actually
getting a btf_fd from sys_bpf(BPF_BTF_GET_FD_BY_ID). However, it
forgot to btf__free() it.
In maps_have_btf() stage, there is no need to test it by really
calling sys_bpf(BPF_BTF_GET_FD_BY_ID). Testing non zero
info.btf_id is good enough.
Also, the err_close case is unnecessary, and also causes double
close() because the calling func do_dump() will close() all fds again.
Fixes: f371215a1ed1 ("bpftool: Match maps by name") Signed-off-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Andrii Nakryiko <andriin@fb.com> Cc: Paul Chaignon <paul.chaignon@orange.com> Link: https://lore.kernel.org/bpf/20200115230019.1101352-1-kafai@fb.com
====================
This patch series introduce batch ops that can be added to bpf maps to
lookup/lookup_and_delete/update/delete more than 1 element at the time,
this is specially useful when syscall overhead is a problem and in case
of hmap it will provide a reliable way of traversing them.
The implementation inclues a generic approach that could potentially be
used by any bpf map and adds it to arraymap, it also includes the specific
implementation of hashmaps which are traversed using buckets instead
of keys.
struct { /* struct used by BPF_MAP_*_BATCH commands */
__aligned_u64 in_batch; /* start batch,
* NULL to start from beginning
*/
__aligned_u64 out_batch; /* output: next start batch */
__aligned_u64 keys;
__aligned_u64 values;
__u32 count; /* input/output:
* input: # of key/value
* elements
* output: # of filled elements
*/
__u32 map_fd;
__u64 elem_flags;
__u64 flags;
} batch;
in_batch and out_batch are only used for lookup and lookup_and_delete since
those are the only two operations that attempt to traverse the map.
update/delete batch ops should provide the keys/values that user wants
to modify.
Here are the previous discussions on the batch processing:
- https://lore.kernel.org/bpf/20190724165803.87470-1-brianvv@google.com/
- https://lore.kernel.org/bpf/20190829064502.2750303-1-yhs@fb.com/
- https://lore.kernel.org/bpf/20190906225434.3635421-1-yhs@fb.com/
Changelog sinve v4:
- Remove unnecessary checks from libbpf API (Andrii Nakryiko)
- Move DECLARE_LIBBPF_OPTS with all var declarations (Andrii Nakryiko)
- Change bucket internal buffer size to 5 entries (Yonghong Song)
- Fix some minor bugs in hashtab batch ops implementation (Yonghong Song)
Changelog sinve v3:
- Do not use copy_to_user inside atomic region (Yonghong Song)
- Use _opts approach on libbpf APIs (Andrii Nakryiko)
- Drop generic_map_lookup_and_delete_batch support
- Free malloc-ed memory in tests (Yonghong Song)
- Reverse christmas tree (Yonghong Song)
- Add acked labels
Changelog sinve v2:
- Add generic batch support for lpm_trie and test it (Yonghong Song)
- Use define MAP_LOOKUP_RETRIES for retries (John Fastabend)
- Return errors directly and remove labels (Yonghong Song)
- Insert new API functions into libbpf alphabetically (Yonghong Song)
- Change hlist_nulls_for_each_entry_rcu to
hlist_nulls_for_each_entry_safe in htab batch ops (Yonghong Song)
Changelog since v1:
- Fix SOB ordering and remove Co-authored-by tag (Alexei Starovoitov)
Changelog since RFC:
- Change batch to in_batch and out_batch to support more flexible opaque
values to iterate the bpf maps.
- Remove update/delete specific batch ops for htab and use the generic
implementations instead.
====================
Acked-by: Yonghong Song <yhs@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Yonghong Song [Wed, 15 Jan 2020 18:43:06 +0000 (10:43 -0800)]
libbpf: Add libbpf support to batch ops
Added four libbpf API functions to support map batch operations:
. int bpf_map_delete_batch( ... )
. int bpf_map_lookup_batch( ... )
. int bpf_map_lookup_and_delete_batch( ... )
. int bpf_map_update_batch( ... )
Yonghong Song [Wed, 15 Jan 2020 18:43:04 +0000 (10:43 -0800)]
bpf: Add batch ops to all htab bpf map
htab can't use generic batch support due some problematic behaviours
inherent to the data structre, i.e. while iterating the bpf map a
concurrent program might delete the next entry that batch was about to
use, in that case there's no easy solution to retrieve the next entry,
the issue has been discussed multiple times (see [1] and [2]).
The only way hmap can be traversed without the problem previously
exposed is by making sure that the map is traversing entire buckets.
This commit implements those strict requirements for hmap, the
implementation follows the same interaction that generic support with
some exceptions:
- If keys/values buffer are not big enough to traverse a bucket,
ENOSPC will be returned.
- out_batch contains the value of the next bucket in the iteration, not
the next key, but this is transparent for the user since the user
should never use out_batch for other than bpf batch syscalls.
This commits implements BPF_MAP_LOOKUP_BATCH and adds support for new
command BPF_MAP_LOOKUP_AND_DELETE_BATCH. Note that for update/delete
batch ops it is possible to use the generic implementations.
Brian Vazquez [Wed, 15 Jan 2020 18:43:03 +0000 (10:43 -0800)]
bpf: Add lookup and update batch ops to arraymap
This adds the generic batch ops functionality to bpf arraymap, note that
since deletion is not a valid operation for arraymap, only batch and
lookup are added.
Brian Vazquez [Wed, 15 Jan 2020 18:43:02 +0000 (10:43 -0800)]
bpf: Add generic support for update and delete batch ops
This commit adds generic support for update and delete batch ops that
can be used for almost all the bpf maps. These commands share the same
UAPI attr that lookup and lookup_and_delete batch ops use and the
syscall commands are:
BPF_MAP_UPDATE_BATCH
BPF_MAP_DELETE_BATCH
The main difference between update/delete and lookup batch ops is that
for update/delete keys/values must be specified for userspace and
because of that, neither in_batch nor out_batch are used.
Brian Vazquez [Wed, 15 Jan 2020 18:43:01 +0000 (10:43 -0800)]
bpf: Add generic support for lookup batch op
This commit introduces generic support for the bpf_map_lookup_batch.
This implementation can be used by almost all the bpf maps since its core
implementation is relying on the existing map_get_next_key and
map_lookup_elem. The bpf syscall subcommand introduced is:
BPF_MAP_LOOKUP_BATCH
The UAPI attribute is:
struct { /* struct used by BPF_MAP_*_BATCH commands */
__aligned_u64 in_batch; /* start batch,
* NULL to start from beginning
*/
__aligned_u64 out_batch; /* output: next start batch */
__aligned_u64 keys;
__aligned_u64 values;
__u32 count; /* input/output:
* input: # of key/value
* elements
* output: # of filled elements
*/
__u32 map_fd;
__u64 elem_flags;
__u64 flags;
} batch;
in_batch/out_batch are opaque values use to communicate between
user/kernel space, in_batch/out_batch must be of key_size length.
To start iterating from the beginning in_batch must be null,
count is the # of key/value elements to retrieve. Note that the 'keys'
buffer must be a buffer of key_size * count size and the 'values' buffer
must be value_size * count, where value_size must be aligned to 8 bytes
by userspace if it's dealing with percpu maps. 'count' will contain the
number of keys/values successfully retrieved. Note that 'count' is an
input/output variable and it can contain a lower value after a call.
If there's no more entries to retrieve, ENOENT will be returned. If error
is ENOENT, count might be > 0 in case it copied some values but there were
no more entries to retrieve.
Note that if the return code is an error and not -EFAULT,
count indicates the number of elements successfully processed.
Eelco Chaudron [Wed, 15 Jan 2020 13:15:39 +0000 (13:15 +0000)]
selftests/bpf: Add a test for attaching a bpf fentry/fexit trace to an XDP program
Add a test that will attach a FENTRY and FEXIT program to the XDP test
program. It will also verify data from the XDP context on FENTRY and
verifies the return code on exit.
Andrii Nakryiko [Wed, 15 Jan 2020 19:08:56 +0000 (11:08 -0800)]
libbpf: Support .text sub-calls relocations
The LLVM patch https://reviews.llvm.org/D72197 makes LLVM emit function call
relocations within the same section. This includes a default .text section,
which contains any BPF sub-programs. This wasn't the case before and so libbpf
was able to get a way with slightly simpler handling of subprogram call
relocations.
This patch adds support for .text section relocations. It needs to ensure
correct order of relocations, so does two passes:
- first, relocate .text instructions, if there are any relocations in it;
- then process all the other programs and copy over patched .text instructions
for all sub-program calls.
v1->v2:
- break early once .text program is processed.
====================
Commit 6947ab335019 ("bpf: implement bpf_send_signal() helper")
added helper bpf_send_signal() which permits bpf program to
send a signal to the current process. The signal may send to
any thread of the process.
This patch implemented a new helper bpf_send_signal_thread()
to send a signal to the thread corresponding to the kernel current task.
This helper can simplify user space code if the thread context of
bpf sending signal is needed in user space. Please see Patch #1 for
details of use case and kernel implementation.
Patch #2 added some bpf self tests for the new helper.
Changelogs:
v2 -> v3:
- More simplification for skeleton codes by removing not-needed
mmap code and redundantly created tracepoint link.
v1 -> v2:
- More description for the difference between bpf_send_signal()
and bpf_send_signal_thread() in the uapi header bpf.h.
- Use skeleton and mmap for send_signal test.
====================
Yonghong Song [Wed, 15 Jan 2020 03:50:02 +0000 (19:50 -0800)]
bpf: Add bpf_send_signal_thread() helper
Commit 6947ab335019 ("bpf: implement bpf_send_signal() helper")
added helper bpf_send_signal() which permits bpf program to
send a signal to the current process. The signal may be
delivered to any threads in the process.
We found a use case where sending the signal to the current
thread is more preferable.
- A bpf program will collect the stack trace and then
send signal to the user application.
- The user application will add some thread specific
information to the just collected stack trace for
later analysis.
If bpf_send_signal() is used, user application will need
to check whether the thread receiving the signal matches
the thread collecting the stack by checking thread id.
If not, it will need to send signal to another thread
through pthread_kill().
This patch proposed a new helper bpf_send_signal_thread(),
which sends the signal to the thread corresponding to
the current kernel task. This way, user space is guaranteed that
bpf_program execution context and user space signal handling
context are the same thread.
Magnus Karlsson [Tue, 14 Jan 2020 09:49:25 +0000 (10:49 +0100)]
xsk: Support allocations of large umems
When registering a umem area that is sufficiently large (>1G on an
x86), kmalloc cannot be used to allocate one of the internal data
structures, as the size requested gets too large. Use kvmalloc instead
that falls back on vmalloc if the allocation is too large for kmalloc.
Also add accounting for this structure as it is triggered by a user
space action (the XDP_UMEM_REG setsockopt) and it is by far the
largest structure of kernel allocated memory in xsk.
Li RongQing [Fri, 10 Jan 2020 01:04:37 +0000 (09:04 +0800)]
bpf: Return -EBADRQC for invalid map type in __bpf_tx_xdp_map
A negative value should be returned if map->map_type is invalid
although that is impossible now, but if we run into such situation
in future, then xdpbuff could be leaked.
Daniel Borkmann suggested:
-EBADRQC should be returned to stay consistent with generic XDP
for the tracepoint output and not to be confused with -EOPNOTSUPP
from other locations like dev_map_enqueue() when ndo_xdp_xmit is
missing and such.
Martin KaFai Lau [Tue, 14 Jan 2020 07:26:47 +0000 (23:26 -0800)]
bpf: Fix seq_show for BPF_MAP_TYPE_STRUCT_OPS
Instead of using bpf_struct_ops_map_lookup_elem() which is
not implemented, bpf_struct_ops_map_seq_show_elem() should
also use bpf_struct_ops_map_sys_lookup_elem() which does
an inplace update to the value. The change allocates
a value to pass to bpf_struct_ops_map_sys_lookup_elem().
====================
Based on recent BPF CO-RE, tp_btf, and BPF skeleton changes, re-implement
BCC-based runqslower tool as a portable pre-compiled BPF CO-RE-based tool.
Make sure it's built as part of selftests to ensure it doesn't bit rot.
As part of this patch set, augment `format c` output of `bpftool btf dump`
sub-command with applying `preserve_access_index` attribute to all structs and
unions. This makes all such structs and unions automatically relocatable under
BPF CO-RE, which improves user experience of writing TRACING programs with
direct kernel memory read access.
Also, further clean up selftest/bpf Makefile output and make it conforming to
libbpf and bpftool succinct output format.
v1->v2:
- build in-tree bpftool for runqslower (Yonghong);
- drop `format core` and augment `format c` instead (Alexei);
- move runqslower under tools/bpf (Daniel).
====================
Andrii Nakryiko [Mon, 13 Jan 2020 07:31:42 +0000 (23:31 -0800)]
tools/bpf: Add runqslower tool to tools/bpf
Convert one of BCC tools (runqslower [0]) to BPF CO-RE + libbpf. It matches
its BCC-based counterpart 1-to-1, supporting all the same parameters and
functionality.
runqslower tool utilizes BPF skeleton, auto-generated from BPF object file,
as well as memory-mapped interface to global (read-only, in this case) data.
Its Makefile also ensures auto-generation of "relocatable" vmlinux.h, which is
necessary for BTF-typed raw tracepoints with direct memory access.
Andrii Nakryiko [Mon, 13 Jan 2020 07:31:41 +0000 (23:31 -0800)]
bpftool: Apply preserve_access_index attribute to all types in BTF dump
This patch makes structs and unions, emitted through BTF dump, automatically
CO-RE-relocatable (unless disabled with `#define BPF_NO_PRESERVE_ACCESS_INDEX`,
specified before including generated header file).
This effectivaly turns usual bpf_probe_read() call into equivalent of
bpf_core_read(), by automatically applying builtin_preserve_access_index to
any field accesses of types in generated C types header.
This is especially useful for tp_btf/fentry/fexit BPF program types. They
allow direct memory access, so BPF C code just uses straightfoward a->b->c
access pattern to read data from kernel. But without kernel structs marked as
CO-RE relocatable through preserve_access_index attribute, one has to enclose
all the data reads into a special __builtin_preserve_access_index code block,
like so:
__builtin_preserve_access_index(({
x = p->pid; /* where p is struct task_struct *, for example */
}));
This is very inconvenient and obscures the logic quite a bit. By marking all
auto-generated types with preserve_access_index attribute the above code is
reduced to just a clean and natural `x = p->pid;`.
Andrii Nakryiko [Mon, 13 Jan 2020 07:31:40 +0000 (23:31 -0800)]
selftests/bpf: Conform selftests/bpf Makefile output to libbpf and bpftool
Bring selftest/bpf's Makefile output to the same format used by libbpf and
bpftool: 2 spaces of padding on the left + 8-character left-aligned build step
identifier.
Also, hide feature detection output by default. Can be enabled back by setting
V=1.
Andrii Nakryiko [Mon, 13 Jan 2020 07:31:39 +0000 (23:31 -0800)]
libbpf: Clean up bpf_helper_defs.h generation output
bpf_helpers_doc.py script, used to generate bpf_helper_defs.h, unconditionally
emits one informational message to stderr. Remove it and preserve stderr to
contain only relevant errors. Also make sure script invocations command is
muted by default in libbpf's Makefile.
Andrii Nakryiko [Fri, 10 Jan 2020 21:16:34 +0000 (13:16 -0800)]
selftests/bpf: Add BPF_PROG, BPF_KPROBE, and BPF_KRETPROBE macros
Streamline BPF_TRACE_x macro by moving out return type and section attribute
definition out of macro itself. That makes those function look in source code
similar to other BPF programs. Additionally, simplify its usage by determining
number of arguments automatically (so just single BPF_TRACE vs a family of
BPF_TRACE_1, BPF_TRACE_2, etc). Also, allow more natural function argument
syntax without commas inbetween argument type and name.
Given this helper is useful not only for tracing tp_btf/fenty/fexit programs,
but could be used for LSM programs and others following the same pattern,
rename BPF_TRACE macro into more generic BPF_PROG. Existing BPF_TRACE_x
usages in selftests are converted to new BPF_PROG macro.
Following the same pattern, define BPF_KPROBE and BPF_KRETPROBE macros for
nicer usage of kprobe/kretprobe arguments, respectively. BPF_KRETPROBE, adopts
same convention used by fexit programs, that last defined argument is probed
function's return result.
v4->v5:
- fix test_overhead test (__set_task_comm is void) (Alexei);
v3->v4:
- rebased and fixed one more BPF_TRACE_x occurence (Alexei);
v2->v3:
- rename to shorter and as generic BPF_PROG (Alexei);
v1->v2:
- verified GCC handles pragmas as expected;
- added descriptions to macros;
- converted new STRUCT_OPS selftest to BPF_HANDLER (worked as expected);
- added original context as 'ctx' parameter, for cases where it has to be
passed into BPF helpers. This might cause an accidental naming collision,
unfortunately, but at least it's easy to work around. Fortunately, this
situation produces quite legible compilation error:
progs/bpf_dctcp.c:46:6: error: redefinition of 'ctx' with a different type: 'int' vs 'unsigned long long *'
int ctx = 123;
^
progs/bpf_dctcp.c:42:6: note: previous definition is here
void BPF_HANDLER(dctcp_init, struct sock *sk)
^
./bpf_trace_helpers.h:58:32: note: expanded from macro 'BPF_HANDLER'
____##name(unsigned long long *ctx, ##args)
Andrii Nakryiko [Fri, 10 Jan 2020 18:19:16 +0000 (10:19 -0800)]
libbpf: Poison kernel-only integer types
It's been a recurring issue with types like u32 slipping into libbpf source
code accidentally. This is not detected during builds inside kernel source
tree, but becomes a compilation error in libbpf's Github repo. Libbpf is
supposed to use only __{s,u}{8,16,32,64} typedefs, so poison {s,u}{8,16,32,64}
explicitly in every .c file. Doing that in a bit more centralized way, e.g.,
inside libbpf_internal.h breaks selftests, which are both using kernel u32 and
libbpf_internal.h.
This patch also fixes a new u32 occurence in libbpf.c, added recently.
Daniel Borkmann [Fri, 10 Jan 2020 16:20:07 +0000 (17:20 +0100)]
Merge branch 'bpf-global-funcs'
Alexei Starovoitov says:
====================
Introduce static vs global functions and function by function verification.
This is another step toward dynamic re-linking (or replacement) of global
functions. See patch 2 for details.
v2->v3:
- cleaned up a check spotted by Song.
- rebased and dropped patch 2 that was trying to improve BTF based on ELF.
- added one more unit test for scalar return value from global func.
v1->v2:
- addressed review comments from Song, Andrii, Yonghong
- fixed memory leak in error path
- added modified ctx check
- added more tests in patch 7
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
selftests/bpf: Add unit tests for global functions
test_global_func[12] - check 512 stack limit.
test_global_func[34] - check 8 frame call chain limit.
test_global_func5 - check that non-ctx pointer cannot be passed into
a function that expects context.
test_global_func6 - check that ctx pointer is unmodified.
test_global_func7 - check that global function returns scalar.
New llvm and old llvm with libbpf help produce BTF that distinguish global and
static functions. Unlike arguments of static function the arguments of global
functions cannot be removed or optimized away by llvm. The compiler has to use
exactly the arguments specified in a function prototype. The argument type
information allows the verifier validate each global function independently.
For now only supported argument types are pointer to context and scalars. In
the future pointers to structures, sizes, pointer to packet data can be
supported as well. Consider the following example:
static int f1(int ...)
{
...
}
int f3(int b);
int f2(int a)
{
f1(a) + f3(a);
}
int f3(int b)
{
...
}
int main(...)
{
f1(...) + f2(...) + f3(...);
}
The verifier will start its safety checks from the first global function f2().
It will recursively descend into f1() because it's static. Then it will check
that arguments match for the f3() invocation inside f2(). It will not descend
into f3(). It will finish f2() that has to be successfully verified for all
possible values of 'a'. Then it will proceed with f3(). That function also has
to be safe for all possible values of 'b'. Then it will start subprog 0 (which
is main() function). It will recursively descend into f1() and will skip full
check of f2() and f3(), since they are global. The order of processing global
functions doesn't affect safety, since all global functions must be proven safe
based on their arguments only.
Such function by function verification can drastically improve speed of the
verification and reduce complexity.
Note that the stack limit of 512 still applies to the call chain regardless whether
functions were static or global. The nested level of 8 also still applies. The
same recursion prevention checks are in place as well.
The type information and static/global kind is preserved after the verification
hence in the above example global function f2() and f3() can be replaced later
by equivalent functions with the same types that are loaded and verified later
without affecting safety of this main() program. Such replacement (re-linking)
of global functions is a subject of future patches.
====================
Fix issues with bpf_helper_defs.h usage in selftests/bpf. As part of that, fix
the way clean up is performed for libbpf and selftests/bpf. Some for Makefile
output clean ups as well.
====================
Andrii Nakryiko [Fri, 10 Jan 2020 05:17:15 +0000 (21:17 -0800)]
selftests/bpf: Ensure bpf_helper_defs.h are taken from selftests dir
Reorder includes search path to ensure $(OUTPUT) and $(CURDIR) go before
libbpf's directory. Also fix bpf_helpers.h to include bpf_helper_defs.h in
such a way as to leverage includes search path. This allows selftests to not
use libbpf's local and potentially stale bpf_helper_defs.h. It's important
because selftests/bpf's Makefile only re-generates bpf_helper_defs.h in
seltests' output directory, not the one in libbpf's directory.
Also force regeneration of bpf_helper_defs.h when libbpf.a is updated to
reduce staleness.
Andrii Nakryiko [Fri, 10 Jan 2020 05:17:14 +0000 (21:17 -0800)]
libbpf,selftests/bpf: Fix clean targets
Libbpf's clean target should clean out generated files in $(OUTPUT) directory
and not make assumption that $(OUTPUT) directory is current working directory.
Selftest's Makefile should delegate cleaning of libbpf-generated files to
libbpf's Makefile. This ensures more robust clean up.
Andrii Nakryiko [Fri, 10 Jan 2020 03:42:46 +0000 (19:42 -0800)]
libbpf: Make bpf_map order and indices stable
Currently, libbpf re-sorts bpf_map structs after all the maps are added and
initialized, which might change their relative order and invalidate any
bpf_map pointer or index taken before that. This is inconvenient and
error-prone. For instance, it can cause .kconfig map index to point to a wrong
map.
Furthermore, libbpf itself doesn't rely on any specific ordering of bpf_maps,
so it's just an unnecessary complication right now. This patch drops sorting
of maps and makes their relative positions fixed. If efficient index is ever
needed, it's better to have a separate array of pointers as a search index,
instead of reordering bpf_map struct in-place. This will be less error-prone
and will allow multiple independent orderings, if necessary (e.g., either by
section index or by name).
Fixes: 0523e889ce08 ("libbpf: Support libbpf-provided extern variables") Reported-by: Martin KaFai Lau <kafai@fb.com> Signed-off-by: Andrii Nakryiko <andriin@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Link: https://lore.kernel.org/bpf/20200110034247.1220142-1-andriin@fb.com
Andrey Ignatov [Wed, 8 Jan 2020 01:40:06 +0000 (17:40 -0800)]
bpf: Document BPF_F_QUERY_EFFECTIVE flag
Document BPF_F_QUERY_EFFECTIVE flag, mostly to clarify how it affects
attach_flags what may not be obvious and what may lead to confision.
Specifically attach_flags is returned only for target_fd but if programs
are inherited from an ancestor cgroup then returned attach_flags for
current cgroup may be confusing. For example, two effective programs of
same attach_type can be returned but w/o BPF_F_ALLOW_MULTI in
attach_flags.
Simple repro:
# bpftool c s /sys/fs/cgroup/path/to/task
ID AttachType AttachFlags Name
# bpftool c s /sys/fs/cgroup/path/to/task effective
ID AttachType AttachFlags Name
95043 ingress tw_ipt_ingress
95048 ingress tw_ingress
====================
This series introduces BPF STRUCT_OPS. It is an infra to allow
implementing some specific kernel's function pointers in BPF.
The first use case included in this series is to implement
TCP congestion control algorithm in BPF (i.e. implement
struct tcp_congestion_ops in BPF).
There has been attempt to move the TCP CC to the user space
(e.g. CCP in TCP). The common arguments are faster turn around,
get away from long-tail kernel versions in production...etc,
which are legit points.
BPF has been the continuous effort to join both kernel and
userspace upsides together (e.g. XDP to gain the performance
advantage without bypassing the kernel). The recent BPF
advancements (in particular BTF-aware verifier, BPF trampoline,
BPF CO-RE...) made implementing kernel struct ops (e.g. tcp cc)
possible in BPF.
The idea is to allow implementing tcp_congestion_ops in bpf.
It allows a faster turnaround for testing algorithm in the
production while leveraging the existing (and continue growing) BPF
feature/framework instead of building one specifically for
userspace TCP CC.
Please see individual patch for details.
The bpftool support will be posted in follow-up patches.
v4:
- Expose tcp_ca_find() to tcp.h in patch 7.
It is used to check the same bpf-tcp-cc
does not exist to guarantee the register()
will succeed.
- set_memory_ro() and then set_memory_x() only after all
trampolines are written to the image in patch 6. (Daniel)
spinlock is replaced by mutex because set_memory_*
requires sleepable context.
v3:
- Fix kbuild error by considering CONFIG_BPF_SYSCALL (kbuild)
- Support anonymous bitfield in patch 4 (Andrii, Yonghong)
- Push boundary safety check to a specific arch's trampoline function
(in patch 6) (Yonghong).
Reuse the WANR_ON_ONCE check in arch_prepare_bpf_trampoline() in x86.
- Check module field is 0 in udata in patch 6 (Yonghong)
- Check zero holes in patch 6 (Andrii)
- s/_btf_vmlinux/btf/ in patch 5 and 7 (Andrii)
- s/check_xxx/is_xxx/ in patch 7 (Andrii)
- Use "struct_ops/" convention in patch 11 (Andrii)
- Use the skel instead of bpf_object in patch 11 (Andrii)
- libbpf: Decide BPF_PROG_TYPE_STRUCT_OPS at open phase by using
find_sec_def()
- libbpf: Avoid a debug message at open phase (Andrii)
- libbpf: Add bpf_program__(is|set)_struct_ops() for consistency (Andrii)
- libbpf: Add "struct_ops" to section_defs (Andrii)
- libbpf: Some code shuffling in init_kern_struct_ops() (Andrii)
- libbpf: A few safety checks (Andrii)
v2:
- Dropped cubic for now. They will be reposted
once there are more clarity in "jiffies" on both
bpf side (about the helper) and
tcp_cubic side (some of jiffies usages are being replaced
by tp->tcp_mstamp)
- Remove unnecssary check on bitfield support from btf_struct_access()
(Yonghong)
- BTF_TYPE_EMIT macro (Yonghong, Andrii)
- value_name's length check to avoid an unlikely
type match during truncation case (Yonghong)
- BUILD_BUG_ON to ensure no trampoline-image overrun
in the future (Yonghong)
- Simplify get_next_key() (Yonghong)
- Added comment to explain how to check mandatory
func ptr in net/ipv4/bpf_tcp_ca.c (Yonghong)
- Rename "__bpf_" to "bpf_struct_ops_" for value prefix (Andrii)
- Add comment to highlight the bpf_dctcp.c is not necessarily
the same as tcp_dctcp.c. (Alexei, Eric)
- libbpf: Renmae "struct_ops" to ".struct_ops" for elf sec (Andrii)
- libbpf: Expose struct_ops as a bpf_map (Andrii)
- libbpf: Support multiple struct_ops in SEC(".struct_ops") (Andrii)
- libbpf: Add bpf_map__attach_struct_ops() (Andrii)
====================
The only sec_name convention is SEC(".struct_ops") to identify the
struct_ops implemented in BPF,
e.g. To implement a tcp_congestion_ops:
SEC(".struct_ops")
struct tcp_congestion_ops dctcp = {
.init = (void *)dctcp_init, /* <-- a bpf_prog */
/* ... some more func prts ... */
.name = "bpf_dctcp",
};
Each struct_ops is defined as a global variable under SEC(".struct_ops")
as above. libbpf creates a map for each variable and the variable name
is the map's name. Multiple struct_ops is supported under
SEC(".struct_ops").
In the bpf_object__open phase, libbpf will look for the SEC(".struct_ops")
section and find out what is the btf-type the struct_ops is
implementing. Note that the btf-type here is referring to
a type in the bpf_prog.o's btf. A "struct bpf_map" is added
by bpf_object__add_map() as other maps do. It will then
collect (through SHT_REL) where are the bpf progs that the
func ptrs are referring to. No btf_vmlinux is needed in
the open phase.
In the bpf_object__load phase, the map-fields, which depend
on the btf_vmlinux, are initialized (in bpf_map__init_kern_struct_ops()).
It will also set the prog->type, prog->attach_btf_id, and
prog->expected_attach_type. Thus, the prog's properties do
not rely on its section name.
[ Currently, the bpf_prog's btf-type ==> btf_vmlinux's btf-type matching
process is as simple as: member-name match + btf-kind match + size match.
If these matching conditions fail, libbpf will reject.
The current targeting support is "struct tcp_congestion_ops" which
most of its members are function pointers.
The member ordering of the bpf_prog's btf-type can be different from
the btf_vmlinux's btf-type. ]
Then, all obj->maps are created as usual (in bpf_object__create_maps()).
Once the maps are created and prog's properties are all set,
the libbpf will proceed to load all the progs.
bpf_map__attach_struct_ops() is added to register a struct_ops
map to a kernel subsystem.
This patch makes "struct tcp_congestion_ops" to be the first user
of BPF STRUCT_OPS. It allows implementing a tcp_congestion_ops
in bpf.
The BPF implemented tcp_congestion_ops can be used like
regular kernel tcp-cc through sysctl and setsockopt. e.g.
[root@arch-fb-vm1 bpf]# sysctl -a | egrep congestion
net.ipv4.tcp_allowed_congestion_control = reno cubic bpf_cubic
net.ipv4.tcp_available_congestion_control = reno bic cubic bpf_cubic
net.ipv4.tcp_congestion_control = bpf_cubic
There has been attempt to move the TCP CC to the user space
(e.g. CCP in TCP). The common arguments are faster turn around,
get away from long-tail kernel versions in production...etc,
which are legit points.
BPF has been the continuous effort to join both kernel and
userspace upsides together (e.g. XDP to gain the performance
advantage without bypassing the kernel). The recent BPF
advancements (in particular BTF-aware verifier, BPF trampoline,
BPF CO-RE...) made implementing kernel struct ops (e.g. tcp cc)
possible in BPF. It allows a faster turnaround for testing algorithm
in the production while leveraging the existing (and continue growing)
BPF feature/framework instead of building one specifically for
userspace TCP CC.
This patch allows write access to a few fields in tcp-sock
(in bpf_tcp_ca_btf_struct_access()).
The optional "get_info" is unsupported now. It can be added
later. One possible way is to output the info with a btf-id
to describe the content.
The patch introduces BPF_MAP_TYPE_STRUCT_OPS. The map value
is a kernel struct with its func ptr implemented in bpf prog.
This new map is the interface to register/unregister/introspect
a bpf implemented kernel struct.
The kernel struct is actually embedded inside another new struct
(or called the "value" struct in the code). For example,
"struct tcp_congestion_ops" is embbeded in:
struct bpf_struct_ops_tcp_congestion_ops {
refcount_t refcnt;
enum bpf_struct_ops_state state;
struct tcp_congestion_ops data; /* <-- kernel subsystem struct here */
}
The map value is "struct bpf_struct_ops_tcp_congestion_ops".
The "bpftool map dump" will then be able to show the
state ("inuse"/"tobefree") and the number of subsystem's refcnt (e.g.
number of tcp_sock in the tcp_congestion_ops case). This "value" struct
is created automatically by a macro. Having a separate "value" struct
will also make extending "struct bpf_struct_ops_XYZ" easier (e.g. adding
"void (*init)(void)" to "struct bpf_struct_ops_XYZ" to do some
initialization works before registering the struct_ops to the kernel
subsystem). The libbpf will take care of finding and populating the
"struct bpf_struct_ops_XYZ" from "struct XYZ".
Register a struct_ops to a kernel subsystem:
1. Load all needed BPF_PROG_TYPE_STRUCT_OPS prog(s)
2. Create a BPF_MAP_TYPE_STRUCT_OPS with attr->btf_vmlinux_value_type_id
set to the btf id "struct bpf_struct_ops_tcp_congestion_ops" of the
running kernel.
Instead of reusing the attr->btf_value_type_id,
btf_vmlinux_value_type_id s added such that attr->btf_fd can still be
used as the "user" btf which could store other useful sysadmin/debug
info that may be introduced in the furture,
e.g. creation-date/compiler-details/map-creator...etc.
3. Create a "struct bpf_struct_ops_tcp_congestion_ops" object as described
in the running kernel btf. Populate the value of this object.
The function ptr should be populated with the prog fds.
4. Call BPF_MAP_UPDATE with the object created in (3) as
the map value. The key is always "0".
During BPF_MAP_UPDATE, the code that saves the kernel-func-ptr's
args as an array of u64 is generated. BPF_MAP_UPDATE also allows
the specific struct_ops to do some final checks in "st_ops->init_member()"
(e.g. ensure all mandatory func ptrs are implemented).
If everything looks good, it will register this kernel struct
to the kernel subsystem. The map will not allow further update
from this point.
Unregister a struct_ops from the kernel subsystem:
BPF_MAP_DELETE with key "0".
Introspect a struct_ops:
BPF_MAP_LOOKUP_ELEM with key "0". The map value returned will
have the prog _id_ populated as the func ptr.
The map value state (enum bpf_struct_ops_state) will transit from:
INIT (map created) =>
INUSE (map updated, i.e. reg) =>
TOBEFREE (map value deleted, i.e. unreg)
The kernel subsystem needs to call bpf_struct_ops_get() and
bpf_struct_ops_put() to manage the "refcnt" in the
"struct bpf_struct_ops_XYZ". This patch uses a separate refcnt
for the purose of tracking the subsystem usage. Another approach
is to reuse the map->refcnt and then "show" (i.e. during map_lookup)
the subsystem's usage by doing map->refcnt - map->usercnt to filter out
the map-fd/pinned-map usage. However, that will also tie down the
future semantics of map->refcnt and map->usercnt.
The very first subsystem's refcnt (during reg()) holds one
count to map->refcnt. When the very last subsystem's refcnt
is gone, it will also release the map->refcnt. All bpf_prog will be
freed when the map->refcnt reaches 0 (i.e. during map_free()).
Misc Notes:
* bpf_struct_ops_map_sys_lookup_elem() is added for syscall lookup.
It does an inplace update on "*value" instead returning a pointer
to syscall.c. Otherwise, it needs a separate copy of "zero" value
for the BPF_STRUCT_OPS_STATE_INIT to avoid races.
* The bpf_struct_ops_map_delete_elem() is also called without
preempt_disable() from map_delete_elem(). It is because
the "->unreg()" may requires sleepable context, e.g.
the "tcp_unregister_congestion_control()".
* "const" is added to some of the existing "struct btf_func_model *"
function arg to avoid a compiler warning caused by this patch.
This patch allows the kernel's struct ops (i.e. func ptr) to be
implemented in BPF. The first use case in this series is the
"struct tcp_congestion_ops" which will be introduced in a
latter patch.
This patch introduces a new prog type BPF_PROG_TYPE_STRUCT_OPS.
The BPF_PROG_TYPE_STRUCT_OPS prog is verified against a particular
func ptr of a kernel struct. The attr->attach_btf_id is the btf id
of a kernel struct. The attr->expected_attach_type is the member
"index" of that kernel struct. The first member of a struct starts
with member index 0. That will avoid ambiguity when a kernel struct
has multiple func ptrs with the same func signature.
For example, a BPF_PROG_TYPE_STRUCT_OPS prog is written
to implement the "init" func ptr of the "struct tcp_congestion_ops".
The attr->attach_btf_id is the btf id of the "struct tcp_congestion_ops"
of the _running_ kernel. The attr->expected_attach_type is 3.
The ctx of BPF_PROG_TYPE_STRUCT_OPS is an array of u64 args saved
by arch_prepare_bpf_trampoline that will be done in the next
patch when introducing BPF_MAP_TYPE_STRUCT_OPS.
"struct bpf_struct_ops" is introduced as a common interface for the kernel
struct that supports BPF_PROG_TYPE_STRUCT_OPS prog. The supporting kernel
struct will need to implement an instance of the "struct bpf_struct_ops".
The supporting kernel struct also needs to implement a bpf_verifier_ops.
During BPF_PROG_LOAD, bpf_struct_ops_find() will find the right
bpf_verifier_ops by searching the attr->attach_btf_id.
A new "btf_struct_access" is also added to the bpf_verifier_ops such
that the supporting kernel struct can optionally provide its own specific
check on accessing the func arg (e.g. provide limited write access).
After btf_vmlinux is parsed, the new bpf_struct_ops_init() is called
to initialize some values (e.g. the btf id of the supporting kernel
struct) and it can only be done once the btf_vmlinux is available.
The R0 checks at BPF_EXIT is excluded for the BPF_PROG_TYPE_STRUCT_OPS prog
if the return type of the prog->aux->attach_func_proto is "void".
bpf: Support bitfield read access in btf_struct_access
This patch allows bitfield access as a scalar.
It checks "off + size > t->size" to avoid accessing bitfield
end up accessing beyond the struct. This check is done
outside of the loop since it is applicable to all access.
It also takes this chance to break early on the "off < moff" case.
selftests/bpf: Restore original comm in test_overhead
test_overhead changes task comm in order to estimate BPF trampoline
overhead but never sets the comm back to the original one.
We have the tests (like core_reloc.c) that have 'test_progs'
as hard-coded expected comm, so let's try to preserve the
original comm.
Currently, everything works because the order of execution is:
first core_recloc, then test_overhead; but let's make it a bit
future-proof.
Other related changes: use 'test_overhead' as new comm instead of
'test' to make it easy to debug and drop '\n' at the end.
Michal Rostecki [Wed, 8 Jan 2020 16:23:53 +0000 (17:23 +0100)]
bpftool: Add misc section and probe for large INSN limit
Introduce a new probe section (misc) for probes not related to concrete
map types, program types, functions or kernel configuration. Introduce a
probe for large INSN limit as the first one in that section.
Example outputs:
# bpftool feature probe
[...]
Scanning miscellaneous eBPF features...
Large program size limit is available
- Remove pipeline, CSR and IRQ because version x.y.z already incorporates
this information.
- Remove bond number because it is not used.
- Remove rev number because register was not implemented, always 0
- Add HW Rev ID register to replace rev number
- Add OTP config select to show the user configuration chosen by
the configurable GPIO pins on start-up
Signed-off-by: Vincent Cheng <vincent.cheng.xh@renesas.com> Signed-off-by: David S. Miller <davem@davemloft.net>
YueHaibing [Tue, 7 Jan 2020 14:14:54 +0000 (22:14 +0800)]
enetc: Fix inconsistent IS_ERR and PTR_ERR
The proper pointer to be passed as argument is hw
Detected using Coccinelle.
Fixes: 4d5b060ed376 ("enetc: Make MDIO accessors more generic and export to include/linux/fsl") Signed-off-by: YueHaibing <yuehaibing@huawei.com> Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Dan Carpenter [Tue, 7 Jan 2020 13:11:43 +0000 (16:11 +0300)]
enetc: Fix an off by one in enetc_setup_tc_txtime()
The priv->tx_ring[] has 16 elements but only priv->num_tx_rings are
set up, the rest are NULL. This ">" comparison should be ">=" to avoid
a potential crash.
Fixes: d79fca2118c7 ("enetc: add support time specific departure base on the qos etf") Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
projects / kernel.git / log