Re: [RFC bpf-next 0/3] bpf: adding map batch processing support

On 11/13/19 2:07 PM, Brian Vazquez wrote:

Hi Yonghong,

Thanks for reviewing it! I'm preparing the changes and will submit them 
sooner.

As for the right way to manage author rights, does anyone know what the 
correct approach is? Should I use Yonghong's patch and apply the 
extended support in different patches (i.e. support per_cpu maps, change 
batch from u64 to __aligned_u64, etc) or it is fine to apply the changes 
in place and write both sign-offs?

The logic flow of the patch set is most important.
You can add me as co-signoff if you reuse significant portion of my code.

Thanks,
Brian

On Tue, Nov 12, 2019 at 6:34 PM Yonghong Song <yhs@fb.com 
<mailto:yhs@fb.com>> wrote:



    On 11/7/19 1:20 PM, Brian Vazquez wrote:
     > This is a follow up in the effort to batch bpf map operations to
    reduce
     > the syscall overhead with the map_ops. I initially proposed the
    idea and
     > the discussion is here:
     >
    https://lore.kernel.org/bpf/20190724165803.87470-1-brianvv@google.com/ (local)
     >
     > Yonghong talked at the LPC about this and also proposed and idea that
     > handles the special weird case of hashtables by doing traversing
    using
     > the buckets as a reference instead of a key. Discussion is here:
     > https://lore.kernel.org/bpf/20190906225434.3635421-1-yhs@fb.com/ (local)
     >
     > This RFC proposes a way to extend batch operations for more data
     > structures by creating generic batch functions that can be used
    instead
     > of implementing the operations for each individual data structure,
     > reducing the code that needs to be maintained. The series
    contains the
     > patches used in Yonghong's RFC and the patch that adds the generic
     > implementation of the operations plus some testing with pcpu hashmaps
     > and arrays. Note that pcpu hashmap shouldn't use the generic
     > implementation and it either should have its own implementation
    or share
     > the one introduced by Yonghong, I added that just as an example
    to show
     > that the generic implementation can be easily added to a data
    structure.
     >
     > What I want to achieve with this RFC is to collect early feedback
    and see if
     > there's any major concern about this before I move forward. I do plan
     > to better separate this into different patches and explain them
    properly
     > in the commit messages.

    Thanks Brian for working on this. The general approach described here
    is good to me. Having a generic implementation for batch operations
    looks good for maps (not hash table, queue/stack, etc.)

     >
     > Current known issues where I would like to discuss are the
    followings:
     >
     > - Because Yonghong's UAPI definition was done specifically for
     >    iterating buckets, the batch field is u64 and is treated as an u64
     >    instead of an opaque pointer, this won't work for other data
    structures
     >    that are going to use a key as a batch token with a size
    greater than
     >    64. Although I think at this point the only key that couldn't be
     >    treated as a u64 is the key of a hashmap, and the hashmap
    won't use
     >    the generic interface.

    The u64 can be changed with a __aligned_u64 opaque value. This way,
    it can represent a pointer or a 64bit value.

     > - Not all the data structures use delete (because it's not a valid
     >    operation) i.e. arrays. So maybe lookup_and_delete_batch
    command is
     >    not needed and we can handle that operation with a
    lookup_batch and a
     >    flag.

    This make sense.

     > - For delete_batch (not just the lookup_and_delete_batch). Is this
     >    operation really needed? If so, shouldn't it be better if the
     >    behaviour is delete the keys provided? I did that with my generic
     >    implementation but Yonghong's delete_batch for a hashmap deletes
     >    buckets.

    We need batched delete in bcc. lookup_and_delete_batch is better as
    it can preserves more new map entries. Alternatively, deleting
    all entries after lookup is another option. But this may remove
    more new map entries. Statistically this may or may not matter though.

    bcc does have a clear_table (clear_map) API, but not clear who is
    using it.

    So, I did not have a concrete use case for delete_batch yet.
    I tend to think we should have delete_batch for API completeness,
    but maybe other people can comment on this as well.

    Maybe initial patch, we can skip it. But we should still ensure
    user interface data structure can handle batch delete if it is
    needed later. The current data structure should handle this
    as far as I know.

     >
     > Brian Vazquez (1):
     >    bpf: add generic batch support
     >
     > Yonghong Song (2):
     >    bpf: adding map batch processing support
     >    tools/bpf: test bpf_map_lookup_and_delete_batch()
     >
     >   include/linux/bpf.h                           |  21 +
     >   include/uapi/linux/bpf.h                      |  22 +
     >   kernel/bpf/arraymap.c                         |   4 +
     >   kernel/bpf/hashtab.c                          | 331 ++++++++++
     >   kernel/bpf/syscall.c                          | 573
    ++++++++++++++----
     >   tools/include/uapi/linux/bpf.h                |  22 +
     >   tools/lib/bpf/bpf.c                           |  59 ++
     >   tools/lib/bpf/bpf.h                           |  13 +
     >   tools/lib/bpf/libbpf.map                      |   4 +
     >   .../map_tests/map_lookup_and_delete_batch.c   | 245 ++++++++
     >   .../map_lookup_and_delete_batch_array.c       | 118 ++++
     >   11 files changed, 1292 insertions(+), 120 deletions(-)
     >   create mode 100644
    tools/testing/selftests/bpf/map_tests/map_lookup_and_delete_batch.c
     >   create mode 100644
    tools/testing/selftests/bpf/map_tests/map_lookup_and_delete_batch_array.c
     >