linux/kernel/bpf/hashtab.c, branch v6.4

bpf: fix a memory leak in the LRU and LRU_PERCPU hash maps

2023-05-22T17:26:39Z

The LRU and LRU_PERCPU maps allocate a new element on update before locking the target hash table bucket. Right after that the maps try to lock the bucket. If this fails, then maps return -EBUSY to the caller without releasing the allocated element. This makes the element untracked: it doesn't belong to either of free lists, and it doesn't belong to the hash table, so can't be re-used; this eventually leads to the permanent -ENOMEM on LRU map updates, which is unexpected. Fix this by returning the element to the local free list if bucket locking fails. Fixes: 20b6cc34ea74 ("bpf: Avoid hashtab deadlock with map_locked") Signed-off-by: Anton Protopopov Link: https://lore.kernel.org/r/20230522154558.2166815-1-aspsk@isovalent.com Signed-off-by: Martin KaFai Lau

bpf: optimize hashmap lookups when key_size is divisible by 4

2023-04-01T22:08:19Z

The BPF hashmap uses the jhash() hash function. There is an optimized version of this hash function which may be used if hash size is a multiple of 4. Apply this optimization to the hashmap in a similar way as it is done in the bloom filter map. On practice the optimization is only noticeable for smaller key sizes, which, however, is sufficient for many applications. An example is listed in the following table of measurements (a hashmap of 65536 elements was used): -------------------------------------------------------------------- | key_size | fullness | lookups /sec | lookups (opt) /sec | gain | -------------------------------------------------------------------- | 4 | 25% | 42.990M | 46.000M | 7.0% | | 4 | 50% | 37.910M | 39.094M | 3.1% | | 4 | 75% | 34.486M | 36.124M | 4.7% | | 4 | 100% | 31.760M | 32.719M | 3.0% | -------------------------------------------------------------------- | 8 | 25% | 43.855M | 49.626M | 13.2% | | 8 | 50% | 38.328M | 42.152M | 10.0% | | 8 | 75% | 34.483M | 38.088M | 10.5% | | 8 | 100% | 31.306M | 34.686M | 10.8% | -------------------------------------------------------------------- | 12 | 25% | 38.398M | 43.770M | 14.0% | | 12 | 50% | 33.336M | 37.712M | 13.1% | | 12 | 75% | 29.917M | 34.440M | 15.1% | | 12 | 100% | 27.322M | 30.480M | 11.6% | -------------------------------------------------------------------- | 16 | 25% | 41.491M | 41.921M | 1.0% | | 16 | 50% | 36.206M | 36.474M | 0.7% | | 16 | 75% | 32.529M | 33.027M | 1.5% | | 16 | 100% | 29.581M | 30.325M | 2.5% | -------------------------------------------------------------------- | 20 | 25% | 34.240M | 36.787M | 7.4% | | 20 | 50% | 30.328M | 32.663M | 7.7% | | 20 | 75% | 27.536M | 29.354M | 6.6% | | 20 | 100% | 24.847M | 26.505M | 6.7% | -------------------------------------------------------------------- | 24 | 25% | 36.329M | 40.608M | 11.8% | | 24 | 50% | 31.444M | 35.059M | 11.5% | | 24 | 75% | 28.426M | 31.452M | 10.6% | | 24 | 100% | 26.278M | 28.741M | 9.4% | -------------------------------------------------------------------- | 28 | 25% | 31.540M | 31.944M | 1.3% | | 28 | 50% | 27.739M | 28.063M | 1.2% | | 28 | 75% | 24.993M | 25.814M | 3.3% | | 28 | 100% | 23.513M | 23.500M | -0.1% | -------------------------------------------------------------------- | 32 | 25% | 32.116M | 33.953M | 5.7% | | 32 | 50% | 28.879M | 29.859M | 3.4% | | 32 | 75% | 26.227M | 26.948M | 2.7% | | 32 | 100% | 23.829M | 24.613M | 3.3% | -------------------------------------------------------------------- | 64 | 25% | 22.535M | 22.554M | 0.1% | | 64 | 50% | 20.471M | 20.675M | 1.0% | | 64 | 75% | 19.077M | 19.146M | 0.4% | | 64 | 100% | 17.710M | 18.131M | 2.4% | -------------------------------------------------------------------- The following script was used to gather the results (SMT & frequency off): cd tools/testing/selftests/bpf for key_size in 4 8 12 16 20 24 28 32 64; do for nr_entries in `seq 16384 16384 65536`; do fullness=$(printf '%3s' $((nr_entries*100/65536))) echo -n "key_size=$key_size: $fullness% full: " sudo ./bench -d2 -a bpf-hashmap-lookup --key_size=$key_size --nr_entries=$nr_entries --max_entries=65536 --nr_loops=2000000 --map_flags=0x40 | grep cpu done echo done Signed-off-by: Anton Protopopov Link: https://lore.kernel.org/r/20230401200602.3275-1-aspsk@isovalent.com Signed-off-by: Alexei Starovoitov

bpf: return long from bpf_map_ops funcs

2023-03-22T22:11:30Z

This patch changes the return types of bpf_map_ops functions to long, where previously int was returned. Using long allows for bpf programs to maintain the sign bit in the absence of sign extension during situations where inlined bpf helper funcs make calls to the bpf_map_ops funcs and a negative error is returned. The definitions of the helper funcs are generated from comments in the bpf uapi header at `include/uapi/linux/bpf.h`. The return type of these helpers was previously changed from int to long in commit bdb7b79b4ce8. For any case where one of the map helpers call the bpf_map_ops funcs that are still returning 32-bit int, a compiler might not include sign extension instructions to properly convert the 32-bit negative value a 64-bit negative value. For example: bpf assembly excerpt of an inlined helper calling a kernel function and checking for a specific error: ; err = bpf_map_update_elem(&mymap, &key, &val, BPF_NOEXIST); ... 46: call 0xffffffffe103291c ; htab_map_update_elem ; if (err && err != -EEXIST) { 4b: cmp $0xffffffffffffffef,%rax ; cmp -EEXIST,%rax kernel function assembly excerpt of return value from `htab_map_update_elem` returning 32-bit int: movl $0xffffffef, %r9d ... movl %r9d, %eax ...results in the comparison: cmp $0xffffffffffffffef, $0x00000000ffffffef Fixes: bdb7b79b4ce8 ("bpf: Switch most helper return values from 32-bit int to 64-bit long") Tested-by: Eduard Zingerman Signed-off-by: JP Kobryn Link: https://lore.kernel.org/r/20230322194754.185781-3-inwardvessel@gmail.com Signed-off-by: Alexei Starovoitov

bpf: hashtab memory usage

2023-03-07T17:33:42Z

htab_map_mem_usage() is introduced to calculate hashmap memory usage. In this helper, some small memory allocations are ignore, as their size is quite small compared with the total size. The inner_map_meta in hash_of_map is also ignored. The result for hashtab as follows, - before this change 1: hash name count_map flags 0x1 <<<< no prealloc, fully set key 16B value 24B max_entries 1048576 memlock 41943040B 2: hash name count_map flags 0x1 <<<< no prealloc, none set key 16B value 24B max_entries 1048576 memlock 41943040B 3: hash name count_map flags 0x0 <<<< prealloc key 16B value 24B max_entries 1048576 memlock 41943040B The memlock is always a fixed size whatever it is preallocated or not, and whatever the count of allocated elements is. - after this change 1: hash name count_map flags 0x1 <<<< non prealloc, fully set key 16B value 24B max_entries 1048576 memlock 117441536B 2: hash name count_map flags 0x1 <<<< non prealloc, non set key 16B value 24B max_entries 1048576 memlock 16778240B 3: hash name count_map flags 0x0 <<<< prealloc key 16B value 24B max_entries 1048576 memlock 109056000B The memlock now is hashtab actually allocated. The result for percpu hash map as follows, - before this change 4: percpu_hash name count_map flags 0x0 <<<< prealloc key 16B value 24B max_entries 1048576 memlock 822083584B 5: percpu_hash name count_map flags 0x1 <<<< no prealloc key 16B value 24B max_entries 1048576 memlock 822083584B - after this change 4: percpu_hash name count_map flags 0x0 key 16B value 24B max_entries 1048576 memlock 897582080B 5: percpu_hash name count_map flags 0x1 key 16B value 24B max_entries 1048576 memlock 922748736B At worst, the difference can be 10x, for example, - before this change 6: hash name count_map flags 0x0 key 4B value 4B max_entries 1048576 memlock 8388608B - after this change 6: hash name count_map flags 0x0 key 4B value 4B max_entries 1048576 memlock 83889408B Signed-off-by: Yafang Shao Acked-by: Hou Tao Link: https://lore.kernel.org/r/20230305124615.12358-4-laoar.shao@gmail.com Signed-off-by: Alexei Starovoitov

bpf: Support kptrs in percpu hashmap and percpu LRU hashmap

2023-03-01T18:24:33Z

Enable support for kptrs in percpu BPF hashmap and percpu BPF LRU hashmap by wiring up the freeing of these kptrs from percpu map elements. Signed-off-by: Kumar Kartikeya Dwivedi Link: https://lore.kernel.org/r/20230225154010.391965-2-memxor@gmail.com Signed-off-by: Alexei Starovoitov

bpf: Zeroing allocated object from slab in bpf memory allocator

2023-02-15T23:40:06Z

Currently the freed element in bpf memory allocator may be immediately reused, for htab map the reuse will reinitialize special fields in map value (e.g., bpf_spin_lock), but lookup procedure may still access these special fields, and it may lead to hard-lockup as shown below: NMI backtrace for cpu 16 CPU: 16 PID: 2574 Comm: htab.bin Tainted: G L 6.1.0+ #1 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), RIP: 0010:queued_spin_lock_slowpath+0x283/0x2c0 ...... Call Trace: copy_map_value_locked+0xb7/0x170 bpf_map_copy_value+0x113/0x3c0 __sys_bpf+0x1c67/0x2780 __x64_sys_bpf+0x1c/0x20 do_syscall_64+0x30/0x60 entry_SYSCALL_64_after_hwframe+0x46/0xb0 ...... For htab map, just like the preallocated case, these is no need to initialize these special fields in map value again once these fields have been initialized. For preallocated htab map, these fields are initialized through __GFP_ZERO in bpf_map_area_alloc(), so do the similar thing for non-preallocated htab in bpf memory allocator. And there is no need to use __GFP_ZERO for per-cpu bpf memory allocator, because __alloc_percpu_gfp() does it implicitly. Fixes: 0fd7c5d43339 ("bpf: Optimize call_rcu in non-preallocated hash map.") Signed-off-by: Hou Tao Link: https://lore.kernel.org/r/20230215082132.3856544-2-houtao@huaweicloud.com Signed-off-by: Alexei Starovoitov

bpf: hash map, avoid deadlock with suitable hash mask

2023-01-13T02:55:42Z

The deadlock still may occur while accessed in NMI and non-NMI context. Because in NMI, we still may access the same bucket but with different map_locked index. For example, on the same CPU, .max_entries = 2, we update the hash map, with key = 4, while running bpf prog in NMI nmi_handle(), to update hash map with key = 20, so it will have the same bucket index but have different map_locked index. To fix this issue, using min mask to hash again. Fixes: 20b6cc34ea74 ("bpf: Avoid hashtab deadlock with map_locked") Signed-off-by: Tonghao Zhang Cc: Alexei Starovoitov Cc: Daniel Borkmann Cc: Andrii Nakryiko Cc: Martin KaFai Lau Cc: Song Liu Cc: Yonghong Song Cc: John Fastabend Cc: KP Singh Cc: Stanislav Fomichev Cc: Hao Luo Cc: Jiri Olsa Cc: Hou Tao Acked-by: Yonghong Song Acked-by: Hou Tao Link: https://lore.kernel.org/r/20230111092903.92389-1-tong@infragraf.org Signed-off-by: Martin KaFai Lau

bpf: Do btf_record_free outside map_free callback

2022-11-18T03:11:31Z

Since the commit being fixed, we now miss freeing btf_record for local storage maps which will have a btf_record populated in case they have bpf_spin_lock element. This was missed because I made the choice of offloading the job to free kptr_off_tab (now btf_record) to the map_free callback when adding support for kptrs. Revisiting the reason for this decision, there is the possibility that the btf_record gets used inside map_free callback (e.g. in case of maps embedding kptrs) to iterate over them and free them, hence doing it before the map_free callback would be leaking special field memory, and do invalid memory access. The btf_record keeps module references which is critical to ensure the dtor call made for referenced kptr is safe to do. If doing it after map_free callback, the map area is already freed, so we cannot access bpf_map structure anymore. To fix this and prevent such lapses in future, move bpf_map_free_record out of the map_free callback, and do it after map_free by remembering the btf_record pointer. There is no need to access bpf_map structure in that case, and we can avoid missing this case when support for new map types is added for other special fields. Since a btf_record and its btf_field_offs are used together, for consistency delay freeing of field_offs as well. While not a problem right now, a lot of code assumes that either both record and field_offs are set or none at once. Note that in case of map of maps (outer maps), inner_map_meta->record is only used during verification, not to free fields in map value, hence we simply keep the bpf_map_free_record call as is in bpf_map_meta_free and never touch map->inner_map_meta in bpf_map_free_deferred. Add a comment making note of these details. Fixes: db559117828d ("bpf: Consolidate spin_lock, timer management into btf_record") Signed-off-by: Kumar Kartikeya Dwivedi Link: https://lore.kernel.org/r/20221118015614.2013203-3-memxor@gmail.com Signed-off-by: Alexei Starovoitov

bpf: Consolidate spin_lock, timer management into btf_record

2022-11-04T05:19:40Z

Now that kptr_off_tab has been refactored into btf_record, and can hold more than one specific field type, accomodate bpf_spin_lock and bpf_timer as well. While they don't require any more metadata than offset, having all special fields in one place allows us to share the same code for allocated user defined types and handle both map values and these allocated objects in a similar fashion. As an optimization, we still keep spin_lock_off and timer_off offsets in the btf_record structure, just to avoid having to find the btf_field struct each time their offset is needed. This is mostly needed to manipulate such objects in a map value at runtime. It's ok to hardcode just one offset as more than one field is disallowed. Signed-off-by: Kumar Kartikeya Dwivedi Link: https://lore.kernel.org/r/20221103191013.1236066-8-memxor@gmail.com Signed-off-by: Alexei Starovoitov

bpf: Refactor kptr_off_tab into btf_record

2022-11-04T04:44:53Z

To prepare the BPF verifier to handle special fields in both map values and program allocated types coming from program BTF, we need to refactor the kptr_off_tab handling code into something more generic and reusable across both cases to avoid code duplication. Later patches also require passing this data to helpers at runtime, so that they can work on user defined types, initialize them, destruct them, etc. The main observation is that both map values and such allocated types point to a type in program BTF, hence they can be handled similarly. We can prepare a field metadata table for both cases and store them in struct bpf_map or struct btf depending on the use case. Hence, refactor the code into generic btf_record and btf_field member structs. The btf_record represents the fields of a specific btf_type in user BTF. The cnt indicates the number of special fields we successfully recognized, and field_mask is a bitmask of fields that were found, to enable quick determination of availability of a certain field. Subsequently, refactor the rest of the code to work with these generic types, remove assumptions about kptr and kptr_off_tab, rename variables to more meaningful names, etc. Signed-off-by: Kumar Kartikeya Dwivedi Link: https://lore.kernel.org/r/20221103191013.1236066-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov