Commit e679654a704e ("bpf: Fix a rcu_sched stall issue with
bpf task/task_file iterator") introduced rate limiting in
bpf_seq_read() to fix a case where traversing too many tasks
and files (tens of millions of files) may cause kernel rcu stall.
But rate limiting won't reduce the amount of work to traverse
all these files.
In practice, for a user process, typically all threads belongs
to that process share the same file table and there is no need
to visit every thread for its files.
This patch implemented a customization for task/task_file iterators
to traverse files only for thread with task->tgid == task->pid,
which will include some kernel threads and user process main threads.
Such reduction of unnecessary work will make iterator runtime
much faster if there are a lot of non-main threads and open
files for the process.
Patch #1 fix an uapi issue for bpf_link_info.iter.
Patch #2 implemented the main_thread_only customization for
task/task_file iterators.
Patch #3 added link_query support for new customization.
Patch #4 added bpftool support and Patch #5 added a selftest.
Yonghong Song (5):
bpf: make bpf_link_info.iter similar to bpf_iter_link_info
bpf: add main_thread_only customization for task/task_file iterators
bpf: add link_query support for newly added main_thread_only info
bpftool: support optional 'task main_thread_only' argument
selftests/bpf: test task_file iterator with main_thread_only
include/linux/bpf.h | 3 +-
include/uapi/linux/bpf.h | 16 ++++-
kernel/bpf/task_iter.c | 63 ++++++++++++++-----
.../bpftool/Documentation/bpftool-iter.rst | 17 ++++-
tools/bpf/bpftool/bash-completion/bpftool | 9 ++-
tools/bpf/bpftool/iter.c | 28 +++++++--
tools/bpf/bpftool/link.c | 12 ++++
tools/include/uapi/linux/bpf.h | 16 ++++-
.../selftests/bpf/prog_tests/bpf_iter.c | 50 +++++++++++----
.../selftests/bpf/progs/bpf_iter_task_file.c | 9 ++-
10 files changed, 183 insertions(+), 40 deletions(-)
--
2.24.1
