[PATCH bpf v3 0/9] Fixes for LPM trie

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



From: Hou Tao <houtao1@xxxxxxxxxx>

Hi,

This patch set fixes several issues for LPM trie. These issues were
found during adding new test cases or were reported by syzbot.

The patch set is structured as follows:

Patch #1~#2 are clean-ups for lpm_trie_update_elem().
Patch #3 handles BPF_EXIST and BPF_NOEXIST correctly for LPM trie.
Patch #4 fixes the accounting of n_entries when doing in-place update.
Patch #5 fixes the exact match condition in trie_get_next_key() and it
may skip keys when the passed key is not found in the map.
Patch #6~#7 switch from kmalloc() to bpf memory allocator for LPM trie
to fix several lock order warnings reported by syzbot. It also enables
raw_spinlock_t for LPM trie again. After these changes, the LPM trie will
be closer to being usable in any context (though the reentrance check of
trie->lock is still missing, but it is on my todo list).
Patch #8: move test_lpm_map to map_tests to make it run regularly.
Patch #9: add test cases for the issues fixed by patch #3~#5.

Please see individual patches for more details. Comments are always
welcome.

Change Log:
v3:
  * patch #2: remove the unnecessary NULL-init for im_node
  * patch #6: alloc the leaf node before disabling IRQ to low
    the possibility of -ENOMEM when leaf_size is large; Free
    these nodes outside the trie lock (Suggested by Alexei)
  * collect review and ack tags (Thanks for Toke & Daniel)

v2: https://lore.kernel.org/bpf/20241127004641.1118269-1-houtao@xxxxxxxxxxxxxxx/
  * collect review tags (Thanks for Toke)
  * drop "Add bpf_mem_cache_is_mergeable() helper" patch
  * patch #3~#4: add fix tag
  * patch #4: rename the helper to trie_check_add_elem() and increase
    n_entries in it.
  * patch #6: use one bpf mem allocator and update commit message to
    clarify that using bpf mem allocator is more appropriate.
  * patch #7: update commit message to add the possible max running time
    for update operation.
  * patch #9: update commit message to specify the purpose of these test
    cases.

v1: https://lore.kernel.org/bpf/20241118010808.2243555-1-houtao@xxxxxxxxxxxxxxx/

Hou Tao (9):
  bpf: Remove unnecessary check when updating LPM trie
  bpf: Remove unnecessary kfree(im_node) in lpm_trie_update_elem
  bpf: Handle BPF_EXIST and BPF_NOEXIST for LPM trie
  bpf: Handle in-place update for full LPM trie correctly
  bpf: Fix exact match conditions in trie_get_next_key()
  bpf: Switch to bpf mem allocator for LPM trie
  bpf: Use raw_spinlock_t for LPM trie
  selftests/bpf: Move test_lpm_map.c to map_tests
  selftests/bpf: Add more test cases for LPM trie

 kernel/bpf/lpm_trie.c                         | 133 +++---
 tools/testing/selftests/bpf/.gitignore        |   1 -
 tools/testing/selftests/bpf/Makefile          |   2 +-
 .../lpm_trie_map_basic_ops.c}                 | 405 +++++++++++++++++-
 4 files changed, 484 insertions(+), 57 deletions(-)
 rename tools/testing/selftests/bpf/{test_lpm_map.c => map_tests/lpm_trie_map_basic_ops.c} (65%)

-- 
2.29.2





[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux