This patch added support for hashmap, percpu hashmap,
lru hashmap and percpu lru hashmap.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210226204927.3885020-1-yhs@fb.com

The bpf_for_each_map_elem() helper is introduced which
iterates all map elements with a callback function. The
helper signature looks like
  long bpf_for_each_map_elem(map, callback_fn, callback_ctx, flags)
and for each map element, the callback_fn will be called. For example,
like hashmap, the callback signature may look like
  long callback_fn(map, key, val, callback_ctx)

There are two known use cases for this. One is from upstream ([1]) where
a for_each_map_elem helper may help implement a timeout mechanism
in a more generic way. Another is from our internal discussion
for a firewall use case where a map contains all the rules. The packet
data can be compared to all these rules to decide allow or deny
the packet.

For array maps, users can already use a bounded loop to traverse
elements. Using this helper can avoid using bounded loop. For other
type of maps (e.g., hash maps) where bounded loop is hard or
impossible to use, this helper provides a convenient way to
operate on all elements.

For callback_fn, besides map and map element, a callback_ctx,
allocated on caller stack, is also passed to the callback
function. This callback_ctx argument can provide additional
input and allow to write to caller stack for output.

If the callback_fn returns 0, the helper will iterate through next
element if available. If the callback_fn returns 1, the helper
will stop iterating and returns to the bpf program. Other return
values are not used for now.

Currently, this helper is only available with jit. It is possible
to make it work with interpreter with so effort but I leave it
as the future work.

[1]: https://lore.kernel.org/bpf/20210122205415.113822-1-xiyou.wangcong@gmail.com/Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210226204925.3884923-1-yhs@fb.com

Currently, verifier function add_subprog() returns 0 for success
and negative value for failure. Change the return value
to be the subprog number for success. This functionality will be
used in the next patch to save a call to find_subprog().
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210226204924.3884848-1-yhs@fb.com

Later proposed bpf_for_each_map_elem() helper has callback
function as one of its arguments. This patch refactored
check_func_call() to permit callback function which sets
callee state. Different callback functions may have
different callee states.
There is no functionality change for this patch.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210226204923.3884627-1-yhs@fb.com

Factor out the function verbose_invalid_scalar() to verbose
print if a scalar is not in a tnum range. There is no
functionality change and the function will be used by
later patch which introduced bpf_for_each_map_elem().
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210226204922.3884375-1-yhs@fb.com

During verifier check_cfg(), all instructions are
visited to ensure verifier can handle program control flows.
This patch factored out function visit_func_call_insn()
so it can be reused in later patch to visit callback function
calls. There is no functionality change for this patch.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210226204920.3884136-1-yhs@fb.com

Building selftests in a separate directory like this:

    make O="$BUILD" -C tools/testing/selftests/bpf

and then running:

    cd "$BUILD" && ./test_progs -t btf

causes all the non-flavored btf_dump_test_case_*.c tests to fail,
because these files are not copied to where test_progs expects to find
them.

Fix by not skipping EXT-COPY when the original $(OUTPUT) is not empty
(lib.mk sets it to $(shell pwd) in that case) and using rsync instead
of cp: cp fails because e.g. urandom_read is being copied into itself,
and rsync simply skips such cases. rsync is already used by kselftests
and therefore is not a new dependency.
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210224111445.102342-1-iii@linux.ibm.com

When vmtest.sh ran a command in a VM, it did not record or propagate the
error code of the command. This made the script less "script-able". The
script now saves the error code of the said command in a file in the VM,
copies the file back to the host and (when available) uses this error
code instead of its own.
Signed-off-by: KP Singh <kpsingh@google.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210225161947.1778590-1-kpsingh@kernel.org

Cong Wang says:

====================

From: Cong Wang <cong.wang@bytedance.com>

This patchset is the first series of patches separated out from
the original large patchset, to make reviews easier. This patchset
does not add any new feature or change any functionality but merely
cleans up the existing sockmap and skmsg code and refactors it, to
prepare for the patches followed up. This passed all BPF selftests.

To see the big picture, the original whole patchset is available
on github: https://github.com/congwang/linux/tree/sockmap

and this patchset is also available on github:
https://github.com/congwang/linux/tree/sockmap1
---
v7: add 1 trivial cleanup patch
    define a mask for sk_redir
    fix CONFIG_BPF_SYSCALL in include/net/udp.h
    make sk_psock_done_strp() static
    move skb_bpf_redirect_clear() to sk_psock_backlog()

v6: fix !CONFIG_INET case

v5: improve CONFIG_BPF_SYSCALL dependency
    add 3 trivial cleanup patches

v4: reuse skb dst instead of skb ext
    fix another Kconfig error

v3: fix a few Kconfig compile errors
    remove an unused variable
    add a comment for bpf_convert_data_end_access()

v2: split the original patchset
    compute data_end with bpf_convert_data_end_access()
    get rid of psock->bpf_running
    reduce the scope of CONFIG_BPF_STREAM_PARSER
    do not add CONFIG_BPF_SOCK_MAP
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

It is not defined or used anywhere.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210223184934.6054-10-xiyou.wangcong@gmail.com

It is now nearly identical to bpf_prog_run_pin_on_cpu() and
it has an unused parameter 'psock', so we can just get rid
of it and call bpf_prog_run_pin_on_cpu() directly.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210223184934.6054-9-xiyou.wangcong@gmail.com

It is only used within skmsg.c so can become static.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210223184934.6054-8-xiyou.wangcong@gmail.com

It is only used within sock_map.c so can become static.
Suggested-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210223184934.6054-7-xiyou.wangcong@gmail.com

These two eBPF programs are tied to BPF_SK_SKB_STREAM_PARSER
and BPF_SK_SKB_STREAM_VERDICT, rename them to reflect the fact
they are only used for TCP. And save the name 'skb_verdict' for
general use later.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Reviewed-by: Lorenz Bauer <lmb@cloudflare.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210223184934.6054-6-xiyou.wangcong@gmail.com

Currently TCP_SKB_CB() is hard-coded in skmsg code, it certainly
does not work for any other non-TCP protocols. We can move them to
skb ext, but it introduces a memory allocation on fast path.

Fortunately, we only need to a word-size to store all the information,
because the flags actually only contains 1 bit so can be just packed
into the lowest bit of the "pointer", which is stored as unsigned
long.

Inside struct sk_buff, '_skb_refdst' can be reused because skb dst is
no longer needed after ->sk_data_ready() so we can just drop it.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210223184934.6054-5-xiyou.wangcong@gmail.com

Currently, we compute ->data_end with a compile-time constant
offset of skb. But as Jakub pointed out, we can actually compute
it in eBPF JIT code at run-time, so that we can competely get
rid of ->data_end. This is similar to skb_shinfo(skb) computation
in bpf_convert_shinfo_access().
Suggested-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210223184934.6054-4-xiyou.wangcong@gmail.com

struct sk_psock_parser is embedded in sk_psock, it is
unnecessary as skb verdict also uses ->saved_data_ready.
We can simply fold these fields into sk_psock, and get rid
of ->enabled.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210223184934.6054-3-xiyou.wangcong@gmail.com

As suggested by John, clean up sockmap related Kconfigs:

Reduce the scope of CONFIG_BPF_STREAM_PARSER down to TCP stream
parser, to reflect its name.

Make the rest sockmap code simply depend on CONFIG_BPF_SYSCALL
and CONFIG_INET, the latter is still needed at this point because
of TCP/UDP proto update. And leave CONFIG_NET_SOCK_MSG untouched,
as it is used by non-sockmap cases.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Reviewed-by: Lorenz Bauer <lmb@cloudflare.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210223184934.6054-2-xiyou.wangcong@gmail.com

Commit 34b2021c ("bpf: Add BPF-helper for MTU checking") added an extra
blank line in bpf helper description. This will make bpf_helpers_doc.py stop
building bpf_helper_defs.h immediately after bpf_check_mtu(), which will
affect future added functions.

Fixes: 34b2021c ("bpf: Add BPF-helper for MTU checking")
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Link: https://lore.kernel.org/bpf/20210223131457.1378978-1-liuhangbin@gmail.comSigned-off-by: Alexei Starovoitov <ast@kernel.org>

Ciara Loftus says:

====================

This series attempts to improve the xsk selftest framework by:
1. making the default output less verbose
2. adding an optional verbose flag to both the test_xsk.sh script and xdpxceiver app.
3. renaming the debug option in the app to to 'dump-pkts' and add a flag to the test_xsk.sh
script which enables the flag in the app.
4. changing how tests are launched - now they are launched from the xdpxceiver app
instead of the script.

Once the improvements are made, a new set of tests are added which test the xsk
statistics.

The output of the test script now looks like:

./test_xsk.sh
PREREQUISITES: [ PASS ]
1..10
ok 1 PASS: SKB NOPOLL
ok 2 PASS: SKB POLL
ok 3 PASS: SKB NOPOLL Socket Teardown
ok 4 PASS: SKB NOPOLL Bi-directional Sockets
ok 5 PASS: SKB NOPOLL Stats
ok 6 PASS: DRV NOPOLL
ok 7 PASS: DRV POLL
ok 8 PASS: DRV NOPOLL Socket Teardown
ok 9 PASS: DRV NOPOLL Bi-directional Sockets
ok 10 PASS: DRV NOPOLL Stats
# Totals: pass:10 fail:0 xfail:0 xpass:0 skip:0 error:0
XSK KSELFTESTS: [ PASS ]

v2->v3:
* Rename dump-pkts to dump_pkts in test_xsk.sh
* Add examples of flag usage to test_xsk.sh

v1->v2:
* Changed '-d' flag in the shell script to '-D' to be consistent with the xdpxceiver app.
* Renamed debug mode to 'dump-pkts' which better reflects the behaviour.
* Use libpf APIs instead of calls to ss for configuring xdp on the links
* Remove mutex init & destroy for each stats test
* Added a description for each of the new statistics tests
* Distinguish between exiting due to initialisation failure vs test failure

This series applies on commit d310ec03

Ciara Loftus (3):
  selftests/bpf: expose and rename debug argument
  selftests/bpf: restructure xsk selftests
  selftests/bpf: introduce xsk statistics tests
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

This commit introduces a range of tests to the xsk testsuite
for validating xsk statistics.

A new test type called 'stats' is added. Within it there are
four sub-tests. Each test configures a scenario which should
trigger the given error statistic. The test passes if the statistic
is successfully incremented.

The four statistics for which tests have been created are:
1. rx dropped
Increase the UMEM frame headroom to a value which results in
insufficient space in the rx buffer for both the packet and the headroom.
2. tx invalid
Set the 'len' field of tx descriptors to an invalid value (umem frame
size + 1).
3. rx ring full
Reduce the size of the RX ring to a fraction of the fill ring size.
4. fill queue empty
Do not populate the fill queue and then try to receive pkts.
Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/bpf/20210223162304.7450-5-ciara.loftus@intel.com

Prior to this commit individual xsk tests were launched from the
shell script 'test_xsk.sh'. When adding a new test type, two new test
configurations had to be added to this file - one for each of the
supported XDP 'modes' (skb or drv). Should zero copy support be added to
the xsk selftest framework in the future, three new test configurations
would need to be added for each new test type. Each new test type also
typically requires new CLI arguments for the xdpxceiver program.

This commit aims to reduce the overhead of adding new tests, by launching
the test configurations from within the xdpxceiver program itself, using
simple loops. Every test is run every time the C program is executed. Many
of the CLI arguments can be removed as a result.
Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/bpf/20210223162304.7450-4-ciara.loftus@intel.com

Launching xdpxceiver with -D enables what was formerly know as 'debug'
mode. Rename this mode to 'dump-pkts' as it better describes the
behavior enabled by the option. New usage:

./xdpxceiver .. -D
or
./xdpxceiver .. --dump-pkts

Also make it possible to pass this flag to the app via the test_xsk.sh
shell script like so:

./test_xsk.sh -D
Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210223162304.7450-3-ciara.loftus@intel.com

Make the xsk tests less verbose by only printing the
essentials. Currently, it is hard to see if the tests passed or not
due to all the printouts. Move the extra printouts to a verbose
option, if further debugging is needed when a problem arises.

To run the xsk tests with verbose output:
./test_xsk.sh -v
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://lore.kernel.org/bpf/20210223162304.7450-2-ciara.loftus@intel.com

This function has become overloaded, it actually does lots of diverse
things in a single pass. Rename it to avoid confusion, and add some
concise commentary.
Signed-off-by: Brendan Jackman <jackmanb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210217104509.2423183-1-jackmanb@google.com

Instead of using integer literal here and there use macro name for
better context.
Signed-off-by: Dmitrii Banshchikov <me@ubique.spb.ru>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20210225202629.585485-1-me@ubique.spb.ru

Song Liu says:

====================

This set enables task local storage for non-BPF_LSM programs.

It is common for tracing BPF program to access per-task data. Currently,
these data are stored in hash tables with pid as the key. In
bcc/libbpftools [1], 9 out of 23 tools use such hash tables. However,
hash table is not ideal for many use case. Task local storage provides
better usability and performance for BPF programs. Please refer to 6/6 for
some performance comparison of task local storage vs. hash table.

Changes v5 => v6:
1. Add inc/dec bpf_task_storage_busy in bpf_local_storage_map_free().

Changes v4 => v5:
1. Fix build w/o CONFIG_NET. (kernel test robot)
2. Remove unnecessary check for !task_storage_ptr(). (Martin)
3. Small changes in commit logs.

Changes v3 => v4:
1. Prevent deadlock from recursive calls of bpf_task_storage_[get|delete].
   (2/6 checks potential deadlock and fails over, 4/6 adds a selftest).

Changes v2 => v3:
1. Make the selftest more robust. (Andrii)
2. Small changes with runqslower. (Andrii)
3. Shortern CC list to make it easy for vger.

Changes v1 => v2:
1. Do not allocate task local storage when the task is being freed.
2. Revise the selftest and added a new test for a task being freed.
3. Minor changes in runqslower.
====================
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Replace hashtab with task local storage in runqslower. This improves the
performance of these BPF programs. The following table summarizes average
runtime of these programs, in nanoseconds:

                          task-local   hash-prealloc   hash-no-prealloc
handle__sched_wakeup             125             340               3124
handle__sched_wakeup_new        2812            1510               2998
handle__sched_switch             151             208                991

Note that, task local storage gives better performance than hashtab for
handle__sched_wakeup and handle__sched_switch. On the other hand, for
handle__sched_wakeup_new, task local storage is slower than hashtab with
prealloc. This is because handle__sched_wakeup_new accesses the data for
the first time, so it has to allocate the data for task local storage.
Once the initial allocation is done, subsequent accesses, as those in
handle__sched_wakeup, are much faster with task local storage. If we
disable hashtab prealloc, task local storage is much faster for all 3
functions.
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210225234319.336131-7-songliubraving@fb.com

Update the Makefile to prefer using $(O)/vmlinux, $(KBUILD_OUTPUT)/vmlinux
(for selftests) or ../../../vmlinux. These two files should have latest
definitions for vmlinux.h.
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210225234319.336131-6-songliubraving@fb.com

Add a test with recursive bpf_task_storage_[get|delete] from fentry
programs on bpf_local_storage_lookup and bpf_local_storage_update. Without
proper deadlock prevent mechanism, this test would cause deadlock.
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210225234319.336131-5-songliubraving@fb.com

Task local storage is enabled for tracing programs. Add two tests for
task local storage without CONFIG_BPF_LSM.

The first test stores a value in sys_enter and read it back in sys_exit.

The second test checks whether the kernel allows allocating task local
storage in exit_creds() (which it should not).
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210225234319.336131-4-songliubraving@fb.com

BPF helpers bpf_task_storage_[get|delete] could hold two locks:
bpf_local_storage_map_bucket->lock and bpf_local_storage->lock. Calling
these helpers from fentry/fexit programs on functions in bpf_*_storage.c
may cause deadlock on either locks.

Prevent such deadlock with a per cpu counter, bpf_task_storage_busy. We
need this counter to be global, because the two locks here belong to two
different objects: bpf_local_storage_map and bpf_local_storage. If we
pick one of them as the owner of the counter, it is still possible to
trigger deadlock on the other lock. For example, if bpf_local_storage_map
owns the counters, it cannot prevent deadlock on bpf_local_storage->lock
when two maps are used.
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20210225234319.336131-3-songliubraving@fb.com

To access per-task data, BPF programs usually creates a hash table with
pid as the key. This is not ideal because:
 1. The user need to estimate the proper size of the hash table, which may
    be inaccurate;
 2. Big hash tables are slow;
 3. To clean up the data properly during task terminations, the user need
    to write extra logic.

Task local storage overcomes these issues and offers a better option for
these per-task data. Task local storage is only available to BPF_LSM. Now
enable it for tracing programs.

Unlike LSM programs, tracing programs can be called in IRQ contexts.
Helpers that access task local storage are updated to use
raw_spin_lock_irqsave() instead of raw_spin_lock_bh().

Tracing programs can attach to functions on the task free path, e.g.
exit_creds(). To avoid allocating task local storage after
bpf_task_storage_free(). bpf_task_storage_get() is updated to not allocate
new storage when the task is not refcounted (task->usage == 0).
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: KP Singh <kpsingh@kernel.org>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20210225234319.336131-2-songliubraving@fb.com

This patch is used to construct skb based on page to save memory copy
overhead.

This function is implemented based on IFF_TX_SKB_NO_LINEAR. Only the
network card priv_flags supports IFF_TX_SKB_NO_LINEAR will use page to
directly construct skb. If this feature is not supported, it is still
necessary to copy data to construct skb.

  ---------------- Performance Testing ------------

  The test environment is Aliyun ECS server.

  Test cmd:
  ```
  xdpsock -i eth0 -t  -S -s <msg size>
  ```

  Test result data:

  size    64      512     1024    1500
  copy    1916747 1775988 1600203 1440054
  page    1974058 1953655 1945463 1904478
  percent 3.0%    10.0%   21.58%  32.3%
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Signed-off-by: Alexander Lobakin <alobakin@pm.me>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Dust Li <dust.li@linux.alibaba.com>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20210218204908.5455-6-alobakin@pm.me

xsk_generic_xmit() allocates a new skb and then queues it for
xmitting. The size of new skb's headroom is desc->len, so it comes
to the driver/device with no reserved headroom and/or tailroom.
Lots of drivers need some headroom (and sometimes tailroom) to
prepend (and/or append) some headers or data, e.g. CPU tags,
device-specific headers/descriptors (LSO, TLS etc.), and if case
of no available space skb_cow_head() will reallocate the skb.
Reallocations are unwanted on fast-path, especially when it comes
to XDP, so generic XSK xmit should reserve the spaces declared in
dev->needed_headroom and dev->needed tailroom to avoid them.

Note on max(NET_SKB_PAD, L1_CACHE_ALIGN(dev->needed_headroom)):

Usually, output functions reserve LL_RESERVED_SPACE(dev), which
consists of dev->hard_header_len + dev->needed_headroom, aligned
by 16.

However, on XSK xmit hard header is already here in the chunk, so
hard_header_len is not needed. But it'd still be better to align
data up to cacheline, while reserving no less than driver requests
for headroom. NET_SKB_PAD here is to double-insure there will be
no reallocations even when the driver advertises no needed_headroom,
but in fact need it (not so rare case).

Fixes: 35fcde7f ("xsk: support for Tx")
Signed-off-by: Alexander Lobakin <alobakin@pm.me>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20210218204908.5455-5-alobakin@pm.me

Virtio net supports the case where the skb linear space is empty, so
add priv_flags.
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Signed-off-by: Alexander Lobakin <alobakin@pm.me>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20210218204908.5455-4-alobakin@pm.me

In some cases, we hope to construct skb directly based on the existing
memory without copying data. In this case, the page will be placed
directly in the skb, and the linear space of skb is empty. But
unfortunately, many the network card does not support this operation.
For example Mellanox Technologies MT27710 Family [ConnectX-4 Lx] will
get the following error message:

    mlx5_core 0000:3b:00.1 eth1: Error cqe on cqn 0x817, ci 0x8,
    qn 0x1dbb, opcode 0xd, syndrome 0x1, vendor syndrome 0x68
    00000000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    00000030: 00 00 00 00 60 10 68 01 0a 00 1d bb 00 0f 9f d2
    WQE DUMP: WQ size 1024 WQ cur size 0, WQE index 0xf, len: 64
    00000000: 00 00 0f 0a 00 1d bb 03 00 00 00 08 00 00 00 00
    00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    00000020: 00 00 00 2b 00 08 00 00 00 00 00 05 9e e3 08 00
    00000030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
    mlx5_core 0000:3b:00.1 eth1: ERR CQE on SQ: 0x1dbb

So a priv_flag is added here to indicate whether the network card
supports this feature.
Suggested-by: Alexander Lobakin <alobakin@pm.me>
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Signed-off-by: Alexander Lobakin <alobakin@pm.me>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20210218204908.5455-3-alobakin@pm.me

This is harmless for now, but can be fatal for future refactors.

Fixes: 871b642a ("netdev: introduce ndo_set_rx_headroom")
Signed-off-by: Alexander Lobakin <alobakin@pm.me>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20210218204908.5455-2-alobakin@pm.me

This adds both the CONFIG_DEBUG_INFO_BTF and CONFIG_DEBUG_INFO_BTF_MODULES
kernel compile option to output of the bpftool feature command.

This is relevant for developers that want to account for data structure
definition differences between kernels.
Signed-off-by: Grant Seltzer <grantseltzer@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20210222195846.155483-1-grantseltzer@gmail.com

Pull performance event updates from Ingo Molnar:

 - Add CPU-PMU support for Intel Sapphire Rapids CPUs

 - Extend the perf ABI with PERF_SAMPLE_WEIGHT_STRUCT, to offer
   two-parameter sampling event feedback. Not used yet, but is intended
   for Golden Cove CPU-PMU, which can provide both the instruction
   latency and the cache latency information for memory profiling
   events.

 - Remove experimental, default-disabled perfmon-v4 counter_freezing
   support that could only be enabled via a boot option. The hardware is
   hopelessly broken, we'd like to make sure nobody starts relying on
   this, as it would only end in tears.

 - Fix energy/power events on Intel SPR platforms

 - Simplify the uprobes resume_execution() logic

 - Misc smaller fixes.

* tag 'perf-core-2021-02-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86/rapl: Fix psys-energy event on Intel SPR platform
  perf/x86/rapl: Only check lower 32bits for RAPL energy counters
  perf/x86/rapl: Add msr mask support
  perf/x86/kvm: Add Cascade Lake Xeon steppings to isolation_ucodes[]
  perf/x86/intel: Support CPUID 10.ECX to disable fixed counters
  perf/x86/intel: Add perf core PMU support for Sapphire Rapids
  perf/x86/intel: Filter unsupported Topdown metrics event
  perf/x86/intel: Factor out intel_update_topdown_event()
  perf/core: Add PERF_SAMPLE_WEIGHT_STRUCT
  perf/intel: Remove Perfmon-v4 counter_freezing support
  x86/perf: Use static_call for x86_pmu.guest_get_msrs
  perf/x86/intel/uncore: With > 8 nodes, get pci bus die id from NUMA info
  perf/x86/intel/uncore: Store the logical die id instead of the physical die id.
  x86/kprobes: Do not decode opcode in resume_execution()