| Message ID | 20201215233702.3301881-1-songliubraving@fb.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | introduce bpf_iter for task_vma | expand |
On 12/15/20 3:36 PM, Song Liu wrote: > This set introduces bpf_iter for task_vma, which can be used to generate > information similar to /proc/pid/maps or /proc/pid/smaps. Patch 4/4 adds I did not see an example for /proc/pid/smaps. It would be good if you can cover smaps as well since it is used by a lot of people. > an example that mimics /proc/pid/maps. > > Changes v1 => v2: > 1. Small fixes in task_iter.c and the selftests. (Yonghong) > > Song Liu (4): > bpf: introduce task_vma bpf_iter > bpf: allow bpf_d_path in sleepable bpf_iter program > libbpf: introduce section "iter.s/" for sleepable bpf_iter program > selftests/bpf: add test for bpf_iter_task_vma > > include/linux/bpf.h | 2 +- > kernel/bpf/task_iter.c | 205 +++++++++++++++++- > kernel/trace/bpf_trace.c | 5 + > tools/lib/bpf/libbpf.c | 5 + > .../selftests/bpf/prog_tests/bpf_iter.c | 106 ++++++++- > tools/testing/selftests/bpf/progs/bpf_iter.h | 9 + > .../selftests/bpf/progs/bpf_iter_task_vma.c | 55 +++++ > 7 files changed, 375 insertions(+), 12 deletions(-) > create mode 100644 tools/testing/selftests/bpf/progs/bpf_iter_task_vma.c > > -- > 2.24.1 >
> On Dec 16, 2020, at 9:00 AM, Yonghong Song <yhs@fb.com> wrote: > > > > On 12/15/20 3:36 PM, Song Liu wrote: >> This set introduces bpf_iter for task_vma, which can be used to generate >> information similar to /proc/pid/maps or /proc/pid/smaps. Patch 4/4 adds > > I did not see an example for /proc/pid/smaps. It would be good if you can cover smaps as well since it is used by a lot of people. smaps is tricky, as it contains a lot of information, and some of these information require architecture and configuration specific logic, e.g., page table structure. To really mimic smaps, we will probably need a helper for smap_gather_stats(). However, I don't think that's really necessary. I think task_vma iterator is most useful in gathering information that are not presented in smaps. For example, if a vma covers mixed 2MB pages and 4kB pages, smaps won't show which address ranges are backed by 2MB pages. I have a test BPF program that parses 4-level x86 page table for huge pages. Since we need bounded-loop to parse the page table, the program won't work well for too big vma. We can probably add this program to samples/bpf/, but I think it is not a good fit for selftests. Thanks, Song
On 12/16/20 9:35 AM, Song Liu wrote: > > >> On Dec 16, 2020, at 9:00 AM, Yonghong Song <yhs@fb.com> wrote: >> >> >> >> On 12/15/20 3:36 PM, Song Liu wrote: >>> This set introduces bpf_iter for task_vma, which can be used to generate >>> information similar to /proc/pid/maps or /proc/pid/smaps. Patch 4/4 adds >> >> I did not see an example for /proc/pid/smaps. It would be good if you can cover smaps as well since it is used by a lot of people. > > smaps is tricky, as it contains a lot of information, and some of these information > require architecture and configuration specific logic, e.g., page table structure. > To really mimic smaps, we will probably need a helper for smap_gather_stats(). > However, I don't think that's really necessary. I think task_vma iterator is most > useful in gathering information that are not presented in smaps. For example, if a > vma covers mixed 2MB pages and 4kB pages, smaps won't show which address ranges are > backed by 2MB pages. Let us remove "/proc/pid/smaps" from cover letter description then. Could you add the above information to the patch #1 (and possibly cover letter as well) of the series? This can serve one of reasons why we introduce task_vma iter. Maybe you want to extend bpf programs to cover this use case. > > I have a test BPF program that parses 4-level x86 page table for huge pages. Since > we need bounded-loop to parse the page table, the program won't work well for too > big vma. We can probably add this program to samples/bpf/, but I think it is not > a good fit for selftests. This can be done later. > > Thanks, > Song >