2014-11-14 09:36:46 +08:00
|
|
|
/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
|
2017-03-16 09:26:42 +08:00
|
|
|
* Copyright (c) 2016,2017 Facebook
|
2014-11-14 09:36:46 +08:00
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of version 2 of the GNU General Public
|
|
|
|
* License as published by the Free Software Foundation.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope that it will be useful, but
|
|
|
|
* WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
|
* General Public License for more details.
|
|
|
|
*/
|
|
|
|
#include <linux/bpf.h>
|
2018-04-19 06:56:03 +08:00
|
|
|
#include <linux/btf.h>
|
2014-11-14 09:36:46 +08:00
|
|
|
#include <linux/err.h>
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/mm.h>
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
#include <linux/filter.h>
|
2015-10-03 00:42:00 +08:00
|
|
|
#include <linux/perf_event.h>
|
2018-04-19 06:56:03 +08:00
|
|
|
#include <uapi/linux/btf.h>
|
2014-11-14 09:36:46 +08:00
|
|
|
|
2017-03-23 01:00:33 +08:00
|
|
|
#include "map_in_map.h"
|
|
|
|
|
2017-10-19 04:00:22 +08:00
|
|
|
#define ARRAY_CREATE_FLAG_MASK \
|
|
|
|
(BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY)
|
|
|
|
|
2016-02-02 14:39:54 +08:00
|
|
|
static void bpf_array_free_percpu(struct bpf_array *array)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
2018-02-23 00:33:24 +08:00
|
|
|
for (i = 0; i < array->map.max_entries; i++) {
|
2016-02-02 14:39:54 +08:00
|
|
|
free_percpu(array->pptrs[i]);
|
2018-02-23 00:33:24 +08:00
|
|
|
cond_resched();
|
|
|
|
}
|
2016-02-02 14:39:54 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static int bpf_array_alloc_percpu(struct bpf_array *array)
|
|
|
|
{
|
|
|
|
void __percpu *ptr;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < array->map.max_entries; i++) {
|
|
|
|
ptr = __alloc_percpu_gfp(array->elem_size, 8,
|
|
|
|
GFP_USER | __GFP_NOWARN);
|
|
|
|
if (!ptr) {
|
|
|
|
bpf_array_free_percpu(array);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
array->pptrs[i] = ptr;
|
2018-02-23 00:33:24 +08:00
|
|
|
cond_resched();
|
2016-02-02 14:39:54 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-11-14 09:36:46 +08:00
|
|
|
/* Called from syscall */
|
2018-01-18 11:13:25 +08:00
|
|
|
static int array_map_alloc_check(union bpf_attr *attr)
|
2014-11-14 09:36:46 +08:00
|
|
|
{
|
2016-02-02 14:39:54 +08:00
|
|
|
bool percpu = attr->map_type == BPF_MAP_TYPE_PERCPU_ARRAY;
|
2017-08-19 02:28:00 +08:00
|
|
|
int numa_node = bpf_map_attr_numa_node(attr);
|
2014-11-14 09:36:46 +08:00
|
|
|
|
|
|
|
/* check sanity of attributes */
|
|
|
|
if (attr->max_entries == 0 || attr->key_size != 4 ||
|
2017-10-19 04:00:22 +08:00
|
|
|
attr->value_size == 0 ||
|
|
|
|
attr->map_flags & ~ARRAY_CREATE_FLAG_MASK ||
|
2017-08-19 02:28:00 +08:00
|
|
|
(percpu && numa_node != NUMA_NO_NODE))
|
2018-01-18 11:13:25 +08:00
|
|
|
return -EINVAL;
|
2014-11-14 09:36:46 +08:00
|
|
|
|
2017-01-11 08:57:30 +08:00
|
|
|
if (attr->value_size > KMALLOC_MAX_SIZE)
|
2015-11-30 08:59:35 +08:00
|
|
|
/* if value_size is bigger, the user space won't be able to
|
|
|
|
* access the elements.
|
|
|
|
*/
|
2018-01-18 11:13:25 +08:00
|
|
|
return -E2BIG;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct bpf_map *array_map_alloc(union bpf_attr *attr)
|
|
|
|
{
|
|
|
|
bool percpu = attr->map_type == BPF_MAP_TYPE_PERCPU_ARRAY;
|
2018-02-16 08:10:29 +08:00
|
|
|
int ret, numa_node = bpf_map_attr_numa_node(attr);
|
2018-01-18 11:13:25 +08:00
|
|
|
u32 elem_size, index_mask, max_entries;
|
|
|
|
bool unpriv = !capable(CAP_SYS_ADMIN);
|
2018-02-16 08:10:29 +08:00
|
|
|
u64 cost, array_size, mask64;
|
2018-01-18 11:13:25 +08:00
|
|
|
struct bpf_array *array;
|
2015-11-30 08:59:35 +08:00
|
|
|
|
2014-11-14 09:36:46 +08:00
|
|
|
elem_size = round_up(attr->value_size, 8);
|
|
|
|
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
max_entries = attr->max_entries;
|
|
|
|
|
bpf, array: fix overflow in max_entries and undefined behavior in index_mask
syzkaller tried to alloc a map with 0xfffffffd entries out of a userns,
and thus unprivileged. With the recently added logic in b2157399cc98
("bpf: prevent out-of-bounds speculation") we round this up to the next
power of two value for max_entries for unprivileged such that we can
apply proper masking into potentially zeroed out map slots.
However, this will generate an index_mask of 0xffffffff, and therefore
a + 1 will let this overflow into new max_entries of 0. This will pass
allocation, etc, and later on map access we still enforce on the original
attr->max_entries value which was 0xfffffffd, therefore triggering GPF
all over the place. Thus bail out on overflow in such case.
Moreover, on 32 bit archs roundup_pow_of_two() can also not be used,
since fls_long(max_entries - 1) can result in 32 and 1UL << 32 in 32 bit
space is undefined. Therefore, do this by hand in a 64 bit variable.
This fixes all the issues triggered by syzkaller's reproducers.
Fixes: b2157399cc98 ("bpf: prevent out-of-bounds speculation")
Reported-by: syzbot+b0efb8e572d01bce1ae0@syzkaller.appspotmail.com
Reported-by: syzbot+6c15e9744f75f2364773@syzkaller.appspotmail.com
Reported-by: syzbot+d2f5524fb46fd3b312ee@syzkaller.appspotmail.com
Reported-by: syzbot+61d23c95395cc90dbc2b@syzkaller.appspotmail.com
Reported-by: syzbot+0d363c942452cca68c01@syzkaller.appspotmail.com
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2018-01-11 06:25:05 +08:00
|
|
|
/* On 32 bit archs roundup_pow_of_two() with max_entries that has
|
|
|
|
* upper most bit set in u32 space is undefined behavior due to
|
|
|
|
* resulting 1U << 32, so do it manually here in u64 space.
|
|
|
|
*/
|
|
|
|
mask64 = fls_long(max_entries - 1);
|
|
|
|
mask64 = 1ULL << mask64;
|
|
|
|
mask64 -= 1;
|
|
|
|
|
|
|
|
index_mask = mask64;
|
|
|
|
if (unpriv) {
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
/* round up array size to nearest power of 2,
|
|
|
|
* since cpu will speculate within index_mask limits
|
|
|
|
*/
|
|
|
|
max_entries = index_mask + 1;
|
bpf, array: fix overflow in max_entries and undefined behavior in index_mask
syzkaller tried to alloc a map with 0xfffffffd entries out of a userns,
and thus unprivileged. With the recently added logic in b2157399cc98
("bpf: prevent out-of-bounds speculation") we round this up to the next
power of two value for max_entries for unprivileged such that we can
apply proper masking into potentially zeroed out map slots.
However, this will generate an index_mask of 0xffffffff, and therefore
a + 1 will let this overflow into new max_entries of 0. This will pass
allocation, etc, and later on map access we still enforce on the original
attr->max_entries value which was 0xfffffffd, therefore triggering GPF
all over the place. Thus bail out on overflow in such case.
Moreover, on 32 bit archs roundup_pow_of_two() can also not be used,
since fls_long(max_entries - 1) can result in 32 and 1UL << 32 in 32 bit
space is undefined. Therefore, do this by hand in a 64 bit variable.
This fixes all the issues triggered by syzkaller's reproducers.
Fixes: b2157399cc98 ("bpf: prevent out-of-bounds speculation")
Reported-by: syzbot+b0efb8e572d01bce1ae0@syzkaller.appspotmail.com
Reported-by: syzbot+6c15e9744f75f2364773@syzkaller.appspotmail.com
Reported-by: syzbot+d2f5524fb46fd3b312ee@syzkaller.appspotmail.com
Reported-by: syzbot+61d23c95395cc90dbc2b@syzkaller.appspotmail.com
Reported-by: syzbot+0d363c942452cca68c01@syzkaller.appspotmail.com
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2018-01-11 06:25:05 +08:00
|
|
|
/* Check for overflows. */
|
|
|
|
if (max_entries < attr->max_entries)
|
|
|
|
return ERR_PTR(-E2BIG);
|
|
|
|
}
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
|
2016-02-02 14:39:54 +08:00
|
|
|
array_size = sizeof(*array);
|
|
|
|
if (percpu)
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
array_size += (u64) max_entries * sizeof(void *);
|
2016-02-02 14:39:54 +08:00
|
|
|
else
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
array_size += (u64) max_entries * elem_size;
|
2016-02-02 14:39:54 +08:00
|
|
|
|
|
|
|
/* make sure there is no u32 overflow later in round_up() */
|
2018-02-16 08:10:29 +08:00
|
|
|
cost = array_size;
|
|
|
|
if (cost >= U32_MAX - PAGE_SIZE)
|
2014-11-19 09:32:16 +08:00
|
|
|
return ERR_PTR(-ENOMEM);
|
2018-02-16 08:10:29 +08:00
|
|
|
if (percpu) {
|
|
|
|
cost += (u64)attr->max_entries * elem_size * num_possible_cpus();
|
|
|
|
if (cost >= U32_MAX - PAGE_SIZE)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
}
|
|
|
|
cost = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT;
|
|
|
|
|
|
|
|
ret = bpf_map_precharge_memlock(cost);
|
|
|
|
if (ret < 0)
|
|
|
|
return ERR_PTR(ret);
|
2014-11-19 09:32:16 +08:00
|
|
|
|
2014-11-14 09:36:46 +08:00
|
|
|
/* allocate all map elements and zero-initialize them */
|
2017-08-19 02:28:00 +08:00
|
|
|
array = bpf_map_area_alloc(array_size, numa_node);
|
bpf: don't trigger OOM killer under pressure with map alloc
This patch adds two helpers, bpf_map_area_alloc() and bpf_map_area_free(),
that are to be used for map allocations. Using kmalloc() for very large
allocations can cause excessive work within the page allocator, so i) fall
back earlier to vmalloc() when the attempt is considered costly anyway,
and even more importantly ii) don't trigger OOM killer with any of the
allocators.
Since this is based on a user space request, for example, when creating
maps with element pre-allocation, we really want such requests to fail
instead of killing other user space processes.
Also, don't spam the kernel log with warnings should any of the allocations
fail under pressure. Given that, we can make backend selection in
bpf_map_area_alloc() generic, and convert all maps over to use this API
for spots with potentially large allocation requests.
Note, replacing the one kmalloc_array() is fine as overflow checks happen
earlier in htab_map_alloc(), since it must also protect the multiplication
for vmalloc() should kmalloc_array() fail.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2017-01-18 22:14:17 +08:00
|
|
|
if (!array)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
array->index_mask = index_mask;
|
|
|
|
array->map.unpriv_array = unpriv;
|
2014-11-14 09:36:46 +08:00
|
|
|
|
|
|
|
/* copy mandatory map attributes */
|
2018-01-18 11:13:26 +08:00
|
|
|
bpf_map_init_from_attr(&array->map, attr);
|
2018-02-16 08:10:29 +08:00
|
|
|
array->map.pages = cost;
|
2014-11-14 09:36:46 +08:00
|
|
|
array->elem_size = elem_size;
|
|
|
|
|
2018-02-16 08:10:29 +08:00
|
|
|
if (percpu && bpf_array_alloc_percpu(array)) {
|
bpf: don't trigger OOM killer under pressure with map alloc
This patch adds two helpers, bpf_map_area_alloc() and bpf_map_area_free(),
that are to be used for map allocations. Using kmalloc() for very large
allocations can cause excessive work within the page allocator, so i) fall
back earlier to vmalloc() when the attempt is considered costly anyway,
and even more importantly ii) don't trigger OOM killer with any of the
allocators.
Since this is based on a user space request, for example, when creating
maps with element pre-allocation, we really want such requests to fail
instead of killing other user space processes.
Also, don't spam the kernel log with warnings should any of the allocations
fail under pressure. Given that, we can make backend selection in
bpf_map_area_alloc() generic, and convert all maps over to use this API
for spots with potentially large allocation requests.
Note, replacing the one kmalloc_array() is fine as overflow checks happen
earlier in htab_map_alloc(), since it must also protect the multiplication
for vmalloc() should kmalloc_array() fail.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2017-01-18 22:14:17 +08:00
|
|
|
bpf_map_area_free(array);
|
2016-02-02 14:39:54 +08:00
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
}
|
|
|
|
|
2014-11-14 09:36:46 +08:00
|
|
|
return &array->map;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Called from syscall or from eBPF program */
|
|
|
|
static void *array_map_lookup_elem(struct bpf_map *map, void *key)
|
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
|
|
|
u32 index = *(u32 *)key;
|
|
|
|
|
2016-02-02 14:39:54 +08:00
|
|
|
if (unlikely(index >= array->map.max_entries))
|
2014-11-14 09:36:46 +08:00
|
|
|
return NULL;
|
|
|
|
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
return array->value + array->elem_size * (index & array->index_mask);
|
2014-11-14 09:36:46 +08:00
|
|
|
}
|
|
|
|
|
2017-03-16 09:26:42 +08:00
|
|
|
/* emit BPF instructions equivalent to C code of array_map_lookup_elem() */
|
|
|
|
static u32 array_map_gen_lookup(struct bpf_map *map, struct bpf_insn *insn_buf)
|
|
|
|
{
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
2017-03-16 09:26:42 +08:00
|
|
|
struct bpf_insn *insn = insn_buf;
|
2017-03-23 01:00:32 +08:00
|
|
|
u32 elem_size = round_up(map->value_size, 8);
|
2017-03-16 09:26:42 +08:00
|
|
|
const int ret = BPF_REG_0;
|
|
|
|
const int map_ptr = BPF_REG_1;
|
|
|
|
const int index = BPF_REG_2;
|
|
|
|
|
|
|
|
*insn++ = BPF_ALU64_IMM(BPF_ADD, map_ptr, offsetof(struct bpf_array, value));
|
|
|
|
*insn++ = BPF_LDX_MEM(BPF_W, ret, index, 0);
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
if (map->unpriv_array) {
|
|
|
|
*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 4);
|
|
|
|
*insn++ = BPF_ALU32_IMM(BPF_AND, ret, array->index_mask);
|
|
|
|
} else {
|
|
|
|
*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 3);
|
|
|
|
}
|
2017-03-23 01:00:32 +08:00
|
|
|
|
|
|
|
if (is_power_of_2(elem_size)) {
|
2017-03-16 09:26:42 +08:00
|
|
|
*insn++ = BPF_ALU64_IMM(BPF_LSH, ret, ilog2(elem_size));
|
|
|
|
} else {
|
|
|
|
*insn++ = BPF_ALU64_IMM(BPF_MUL, ret, elem_size);
|
|
|
|
}
|
|
|
|
*insn++ = BPF_ALU64_REG(BPF_ADD, ret, map_ptr);
|
|
|
|
*insn++ = BPF_JMP_IMM(BPF_JA, 0, 0, 1);
|
|
|
|
*insn++ = BPF_MOV64_IMM(ret, 0);
|
|
|
|
return insn - insn_buf;
|
|
|
|
}
|
|
|
|
|
2016-02-02 14:39:54 +08:00
|
|
|
/* Called from eBPF program */
|
|
|
|
static void *percpu_array_map_lookup_elem(struct bpf_map *map, void *key)
|
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
|
|
|
u32 index = *(u32 *)key;
|
|
|
|
|
|
|
|
if (unlikely(index >= array->map.max_entries))
|
|
|
|
return NULL;
|
|
|
|
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
return this_cpu_ptr(array->pptrs[index & array->index_mask]);
|
2016-02-02 14:39:54 +08:00
|
|
|
}
|
|
|
|
|
bpf: add lookup/update support for per-cpu hash and array maps
The functions bpf_map_lookup_elem(map, key, value) and
bpf_map_update_elem(map, key, value, flags) need to get/set
values from all-cpus for per-cpu hash and array maps,
so that user space can aggregate/update them as necessary.
Example of single counter aggregation in user space:
unsigned int nr_cpus = sysconf(_SC_NPROCESSORS_CONF);
long values[nr_cpus];
long value = 0;
bpf_lookup_elem(fd, key, values);
for (i = 0; i < nr_cpus; i++)
value += values[i];
The user space must provide round_up(value_size, 8) * nr_cpus
array to get/set values, since kernel will use 'long' copy
of per-cpu values to try to copy good counters atomically.
It's a best-effort, since bpf programs and user space are racing
to access the same memory.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-02 14:39:55 +08:00
|
|
|
int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value)
|
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
|
|
|
u32 index = *(u32 *)key;
|
|
|
|
void __percpu *pptr;
|
|
|
|
int cpu, off = 0;
|
|
|
|
u32 size;
|
|
|
|
|
|
|
|
if (unlikely(index >= array->map.max_entries))
|
|
|
|
return -ENOENT;
|
|
|
|
|
|
|
|
/* per_cpu areas are zero-filled and bpf programs can only
|
|
|
|
* access 'value_size' of them, so copying rounded areas
|
|
|
|
* will not leak any kernel data
|
|
|
|
*/
|
|
|
|
size = round_up(map->value_size, 8);
|
|
|
|
rcu_read_lock();
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
pptr = array->pptrs[index & array->index_mask];
|
bpf: add lookup/update support for per-cpu hash and array maps
The functions bpf_map_lookup_elem(map, key, value) and
bpf_map_update_elem(map, key, value, flags) need to get/set
values from all-cpus for per-cpu hash and array maps,
so that user space can aggregate/update them as necessary.
Example of single counter aggregation in user space:
unsigned int nr_cpus = sysconf(_SC_NPROCESSORS_CONF);
long values[nr_cpus];
long value = 0;
bpf_lookup_elem(fd, key, values);
for (i = 0; i < nr_cpus; i++)
value += values[i];
The user space must provide round_up(value_size, 8) * nr_cpus
array to get/set values, since kernel will use 'long' copy
of per-cpu values to try to copy good counters atomically.
It's a best-effort, since bpf programs and user space are racing
to access the same memory.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-02 14:39:55 +08:00
|
|
|
for_each_possible_cpu(cpu) {
|
|
|
|
bpf_long_memcpy(value + off, per_cpu_ptr(pptr, cpu), size);
|
|
|
|
off += size;
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-11-14 09:36:46 +08:00
|
|
|
/* Called from syscall */
|
|
|
|
static int array_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
|
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
2017-04-25 10:00:37 +08:00
|
|
|
u32 index = key ? *(u32 *)key : U32_MAX;
|
2014-11-14 09:36:46 +08:00
|
|
|
u32 *next = (u32 *)next_key;
|
|
|
|
|
|
|
|
if (index >= array->map.max_entries) {
|
|
|
|
*next = 0;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (index == array->map.max_entries - 1)
|
|
|
|
return -ENOENT;
|
|
|
|
|
|
|
|
*next = index + 1;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Called from syscall or from eBPF program */
|
|
|
|
static int array_map_update_elem(struct bpf_map *map, void *key, void *value,
|
|
|
|
u64 map_flags)
|
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
|
|
|
u32 index = *(u32 *)key;
|
|
|
|
|
2016-02-02 14:39:54 +08:00
|
|
|
if (unlikely(map_flags > BPF_EXIST))
|
2014-11-14 09:36:46 +08:00
|
|
|
/* unknown flags */
|
|
|
|
return -EINVAL;
|
|
|
|
|
2016-02-02 14:39:54 +08:00
|
|
|
if (unlikely(index >= array->map.max_entries))
|
2014-11-14 09:36:46 +08:00
|
|
|
/* all elements were pre-allocated, cannot insert a new one */
|
|
|
|
return -E2BIG;
|
|
|
|
|
2016-02-02 14:39:54 +08:00
|
|
|
if (unlikely(map_flags == BPF_NOEXIST))
|
2014-11-19 09:32:16 +08:00
|
|
|
/* all elements already exist */
|
2014-11-14 09:36:46 +08:00
|
|
|
return -EEXIST;
|
|
|
|
|
2016-02-02 14:39:54 +08:00
|
|
|
if (array->map.map_type == BPF_MAP_TYPE_PERCPU_ARRAY)
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
memcpy(this_cpu_ptr(array->pptrs[index & array->index_mask]),
|
2016-02-02 14:39:54 +08:00
|
|
|
value, map->value_size);
|
|
|
|
else
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
memcpy(array->value +
|
|
|
|
array->elem_size * (index & array->index_mask),
|
2016-02-02 14:39:54 +08:00
|
|
|
value, map->value_size);
|
2014-11-14 09:36:46 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
bpf: add lookup/update support for per-cpu hash and array maps
The functions bpf_map_lookup_elem(map, key, value) and
bpf_map_update_elem(map, key, value, flags) need to get/set
values from all-cpus for per-cpu hash and array maps,
so that user space can aggregate/update them as necessary.
Example of single counter aggregation in user space:
unsigned int nr_cpus = sysconf(_SC_NPROCESSORS_CONF);
long values[nr_cpus];
long value = 0;
bpf_lookup_elem(fd, key, values);
for (i = 0; i < nr_cpus; i++)
value += values[i];
The user space must provide round_up(value_size, 8) * nr_cpus
array to get/set values, since kernel will use 'long' copy
of per-cpu values to try to copy good counters atomically.
It's a best-effort, since bpf programs and user space are racing
to access the same memory.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-02 14:39:55 +08:00
|
|
|
int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
|
|
|
|
u64 map_flags)
|
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
|
|
|
u32 index = *(u32 *)key;
|
|
|
|
void __percpu *pptr;
|
|
|
|
int cpu, off = 0;
|
|
|
|
u32 size;
|
|
|
|
|
|
|
|
if (unlikely(map_flags > BPF_EXIST))
|
|
|
|
/* unknown flags */
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (unlikely(index >= array->map.max_entries))
|
|
|
|
/* all elements were pre-allocated, cannot insert a new one */
|
|
|
|
return -E2BIG;
|
|
|
|
|
|
|
|
if (unlikely(map_flags == BPF_NOEXIST))
|
|
|
|
/* all elements already exist */
|
|
|
|
return -EEXIST;
|
|
|
|
|
|
|
|
/* the user space will provide round_up(value_size, 8) bytes that
|
|
|
|
* will be copied into per-cpu area. bpf programs can only access
|
|
|
|
* value_size of it. During lookup the same extra bytes will be
|
|
|
|
* returned or zeros which were zero-filled by percpu_alloc,
|
|
|
|
* so no kernel data leaks possible
|
|
|
|
*/
|
|
|
|
size = round_up(map->value_size, 8);
|
|
|
|
rcu_read_lock();
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
pptr = array->pptrs[index & array->index_mask];
|
bpf: add lookup/update support for per-cpu hash and array maps
The functions bpf_map_lookup_elem(map, key, value) and
bpf_map_update_elem(map, key, value, flags) need to get/set
values from all-cpus for per-cpu hash and array maps,
so that user space can aggregate/update them as necessary.
Example of single counter aggregation in user space:
unsigned int nr_cpus = sysconf(_SC_NPROCESSORS_CONF);
long values[nr_cpus];
long value = 0;
bpf_lookup_elem(fd, key, values);
for (i = 0; i < nr_cpus; i++)
value += values[i];
The user space must provide round_up(value_size, 8) * nr_cpus
array to get/set values, since kernel will use 'long' copy
of per-cpu values to try to copy good counters atomically.
It's a best-effort, since bpf programs and user space are racing
to access the same memory.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-02-02 14:39:55 +08:00
|
|
|
for_each_possible_cpu(cpu) {
|
|
|
|
bpf_long_memcpy(per_cpu_ptr(pptr, cpu), value + off, size);
|
|
|
|
off += size;
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2014-11-14 09:36:46 +08:00
|
|
|
/* Called from syscall or from eBPF program */
|
|
|
|
static int array_map_delete_elem(struct bpf_map *map, void *key)
|
|
|
|
{
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Called when map->refcnt goes to zero, either from workqueue or from syscall */
|
|
|
|
static void array_map_free(struct bpf_map *map)
|
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
|
|
|
|
|
|
|
/* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
|
|
|
|
* so the programs (can be more than one that used this map) were
|
|
|
|
* disconnected from events. Wait for outstanding programs to complete
|
|
|
|
* and free the array
|
|
|
|
*/
|
|
|
|
synchronize_rcu();
|
|
|
|
|
2016-02-02 14:39:54 +08:00
|
|
|
if (array->map.map_type == BPF_MAP_TYPE_PERCPU_ARRAY)
|
|
|
|
bpf_array_free_percpu(array);
|
|
|
|
|
bpf: don't trigger OOM killer under pressure with map alloc
This patch adds two helpers, bpf_map_area_alloc() and bpf_map_area_free(),
that are to be used for map allocations. Using kmalloc() for very large
allocations can cause excessive work within the page allocator, so i) fall
back earlier to vmalloc() when the attempt is considered costly anyway,
and even more importantly ii) don't trigger OOM killer with any of the
allocators.
Since this is based on a user space request, for example, when creating
maps with element pre-allocation, we really want such requests to fail
instead of killing other user space processes.
Also, don't spam the kernel log with warnings should any of the allocations
fail under pressure. Given that, we can make backend selection in
bpf_map_area_alloc() generic, and convert all maps over to use this API
for spots with potentially large allocation requests.
Note, replacing the one kmalloc_array() is fine as overflow checks happen
earlier in htab_map_alloc(), since it must also protect the multiplication
for vmalloc() should kmalloc_array() fail.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2017-01-18 22:14:17 +08:00
|
|
|
bpf_map_area_free(array);
|
2014-11-14 09:36:46 +08:00
|
|
|
}
|
|
|
|
|
2018-04-19 06:56:03 +08:00
|
|
|
static void array_map_seq_show_elem(struct bpf_map *map, void *key,
|
|
|
|
struct seq_file *m)
|
|
|
|
{
|
|
|
|
void *value;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
|
|
|
|
value = array_map_lookup_elem(map, key);
|
|
|
|
if (!value) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
seq_printf(m, "%u: ", *(u32 *)key);
|
2018-05-23 05:57:21 +08:00
|
|
|
btf_type_seq_show(map->btf, map->btf_value_type_id, value, m);
|
2018-04-19 06:56:03 +08:00
|
|
|
seq_puts(m, "\n");
|
|
|
|
|
|
|
|
rcu_read_unlock();
|
|
|
|
}
|
|
|
|
|
|
|
|
static int array_map_check_btf(const struct bpf_map *map, const struct btf *btf,
|
|
|
|
u32 btf_key_id, u32 btf_value_id)
|
|
|
|
{
|
|
|
|
const struct btf_type *key_type, *value_type;
|
|
|
|
u32 key_size, value_size;
|
|
|
|
u32 int_data;
|
|
|
|
|
|
|
|
key_type = btf_type_id_size(btf, &btf_key_id, &key_size);
|
|
|
|
if (!key_type || BTF_INFO_KIND(key_type->info) != BTF_KIND_INT)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
int_data = *(u32 *)(key_type + 1);
|
|
|
|
/* bpf array can only take a u32 key. This check makes
|
|
|
|
* sure that the btf matches the attr used during map_create.
|
|
|
|
*/
|
|
|
|
if (BTF_INT_BITS(int_data) != 32 || key_size != 4 ||
|
|
|
|
BTF_INT_OFFSET(int_data))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
value_type = btf_type_id_size(btf, &btf_value_id, &value_size);
|
|
|
|
if (!value_type || value_size > map->value_size)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-04-11 21:34:58 +08:00
|
|
|
const struct bpf_map_ops array_map_ops = {
|
2018-01-18 11:13:25 +08:00
|
|
|
.map_alloc_check = array_map_alloc_check,
|
2014-11-14 09:36:46 +08:00
|
|
|
.map_alloc = array_map_alloc,
|
|
|
|
.map_free = array_map_free,
|
|
|
|
.map_get_next_key = array_map_get_next_key,
|
|
|
|
.map_lookup_elem = array_map_lookup_elem,
|
|
|
|
.map_update_elem = array_map_update_elem,
|
|
|
|
.map_delete_elem = array_map_delete_elem,
|
2017-03-16 09:26:42 +08:00
|
|
|
.map_gen_lookup = array_map_gen_lookup,
|
2018-04-19 06:56:03 +08:00
|
|
|
.map_seq_show_elem = array_map_seq_show_elem,
|
|
|
|
.map_check_btf = array_map_check_btf,
|
2014-11-14 09:36:46 +08:00
|
|
|
};
|
|
|
|
|
2017-04-11 21:34:58 +08:00
|
|
|
const struct bpf_map_ops percpu_array_map_ops = {
|
2018-01-18 11:13:25 +08:00
|
|
|
.map_alloc_check = array_map_alloc_check,
|
2016-02-02 14:39:54 +08:00
|
|
|
.map_alloc = array_map_alloc,
|
|
|
|
.map_free = array_map_free,
|
|
|
|
.map_get_next_key = array_map_get_next_key,
|
|
|
|
.map_lookup_elem = percpu_array_map_lookup_elem,
|
|
|
|
.map_update_elem = array_map_update_elem,
|
|
|
|
.map_delete_elem = array_map_delete_elem,
|
|
|
|
};
|
|
|
|
|
2018-01-18 11:13:25 +08:00
|
|
|
static int fd_array_map_alloc_check(union bpf_attr *attr)
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
{
|
2015-08-06 15:02:33 +08:00
|
|
|
/* only file descriptors can be stored in this type of map */
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
if (attr->value_size != sizeof(u32))
|
2018-01-18 11:13:25 +08:00
|
|
|
return -EINVAL;
|
|
|
|
return array_map_alloc_check(attr);
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
}
|
|
|
|
|
2015-08-06 15:02:33 +08:00
|
|
|
static void fd_array_map_free(struct bpf_map *map)
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
|
|
|
int i;
|
|
|
|
|
|
|
|
synchronize_rcu();
|
|
|
|
|
|
|
|
/* make sure it's empty */
|
|
|
|
for (i = 0; i < array->map.max_entries; i++)
|
2015-08-06 15:02:33 +08:00
|
|
|
BUG_ON(array->ptrs[i] != NULL);
|
bpf: don't trigger OOM killer under pressure with map alloc
This patch adds two helpers, bpf_map_area_alloc() and bpf_map_area_free(),
that are to be used for map allocations. Using kmalloc() for very large
allocations can cause excessive work within the page allocator, so i) fall
back earlier to vmalloc() when the attempt is considered costly anyway,
and even more importantly ii) don't trigger OOM killer with any of the
allocators.
Since this is based on a user space request, for example, when creating
maps with element pre-allocation, we really want such requests to fail
instead of killing other user space processes.
Also, don't spam the kernel log with warnings should any of the allocations
fail under pressure. Given that, we can make backend selection in
bpf_map_area_alloc() generic, and convert all maps over to use this API
for spots with potentially large allocation requests.
Note, replacing the one kmalloc_array() is fine as overflow checks happen
earlier in htab_map_alloc(), since it must also protect the multiplication
for vmalloc() should kmalloc_array() fail.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2017-01-18 22:14:17 +08:00
|
|
|
|
|
|
|
bpf_map_area_free(array);
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
}
|
|
|
|
|
2015-08-06 15:02:33 +08:00
|
|
|
static void *fd_array_map_lookup_elem(struct bpf_map *map, void *key)
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
{
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2017-06-28 14:08:34 +08:00
|
|
|
/* only called from syscall */
|
|
|
|
int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value)
|
|
|
|
{
|
|
|
|
void **elem, *ptr;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
if (!map->ops->map_fd_sys_lookup_elem)
|
|
|
|
return -ENOTSUPP;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
elem = array_map_lookup_elem(map, key);
|
|
|
|
if (elem && (ptr = READ_ONCE(*elem)))
|
|
|
|
*value = map->ops->map_fd_sys_lookup_elem(ptr);
|
|
|
|
else
|
|
|
|
ret = -ENOENT;
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
/* only called from syscall */
|
2016-06-16 04:47:13 +08:00
|
|
|
int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
|
|
|
|
void *key, void *value, u64 map_flags)
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
2015-08-06 15:02:33 +08:00
|
|
|
void *new_ptr, *old_ptr;
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
u32 index = *(u32 *)key, ufd;
|
|
|
|
|
|
|
|
if (map_flags != BPF_ANY)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (index >= array->map.max_entries)
|
|
|
|
return -E2BIG;
|
|
|
|
|
|
|
|
ufd = *(u32 *)value;
|
2016-06-16 04:47:13 +08:00
|
|
|
new_ptr = map->ops->map_fd_get_ptr(map, map_file, ufd);
|
2015-08-06 15:02:33 +08:00
|
|
|
if (IS_ERR(new_ptr))
|
|
|
|
return PTR_ERR(new_ptr);
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
|
2015-08-06 15:02:33 +08:00
|
|
|
old_ptr = xchg(array->ptrs + index, new_ptr);
|
|
|
|
if (old_ptr)
|
|
|
|
map->ops->map_fd_put_ptr(old_ptr);
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2015-08-06 15:02:33 +08:00
|
|
|
static int fd_array_map_delete_elem(struct bpf_map *map, void *key)
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
2015-08-06 15:02:33 +08:00
|
|
|
void *old_ptr;
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
u32 index = *(u32 *)key;
|
|
|
|
|
|
|
|
if (index >= array->map.max_entries)
|
|
|
|
return -E2BIG;
|
|
|
|
|
2015-08-06 15:02:33 +08:00
|
|
|
old_ptr = xchg(array->ptrs + index, NULL);
|
|
|
|
if (old_ptr) {
|
|
|
|
map->ops->map_fd_put_ptr(old_ptr);
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
return 0;
|
|
|
|
} else {
|
|
|
|
return -ENOENT;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-06-16 04:47:13 +08:00
|
|
|
static void *prog_fd_array_get_ptr(struct bpf_map *map,
|
|
|
|
struct file *map_file, int fd)
|
2015-08-06 15:02:33 +08:00
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
|
|
|
struct bpf_prog *prog = bpf_prog_get(fd);
|
2016-06-16 04:47:13 +08:00
|
|
|
|
2015-08-06 15:02:33 +08:00
|
|
|
if (IS_ERR(prog))
|
|
|
|
return prog;
|
|
|
|
|
|
|
|
if (!bpf_prog_array_compatible(array, prog)) {
|
|
|
|
bpf_prog_put(prog);
|
|
|
|
return ERR_PTR(-EINVAL);
|
|
|
|
}
|
2016-06-16 04:47:13 +08:00
|
|
|
|
2015-08-06 15:02:33 +08:00
|
|
|
return prog;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void prog_fd_array_put_ptr(void *ptr)
|
|
|
|
{
|
bpf: generally move prog destruction to RCU deferral
Jann Horn reported following analysis that could potentially result
in a very hard to trigger (if not impossible) UAF race, to quote his
event timeline:
- Set up a process with threads T1, T2 and T3
- Let T1 set up a socket filter F1 that invokes another filter F2
through a BPF map [tail call]
- Let T1 trigger the socket filter via a unix domain socket write,
don't wait for completion
- Let T2 call PERF_EVENT_IOC_SET_BPF with F2, don't wait for completion
- Now T2 should be behind bpf_prog_get(), but before bpf_prog_put()
- Let T3 close the file descriptor for F2, dropping the reference
count of F2 to 2
- At this point, T1 should have looked up F2 from the map, but not
finished executing it
- Let T3 remove F2 from the BPF map, dropping the reference count of
F2 to 1
- Now T2 should call bpf_prog_put() (wrong BPF program type), dropping
the reference count of F2 to 0 and scheduling bpf_prog_free_deferred()
via schedule_work()
- At this point, the BPF program could be freed
- BPF execution is still running in a freed BPF program
While at PERF_EVENT_IOC_SET_BPF time it's only guaranteed that the perf
event fd we're doing the syscall on doesn't disappear from underneath us
for whole syscall time, it may not be the case for the bpf fd used as
an argument only after we did the put. It needs to be a valid fd pointing
to a BPF program at the time of the call to make the bpf_prog_get() and
while T2 gets preempted, F2 must have dropped reference to 1 on the other
CPU. The fput() from the close() in T3 should also add additionally delay
to the reference drop via exit_task_work() when bpf_prog_release() gets
called as well as scheduling bpf_prog_free_deferred().
That said, it makes nevertheless sense to move the BPF prog destruction
generally after RCU grace period to guarantee that such scenario above,
but also others as recently fixed in ceb56070359b ("bpf, perf: delay release
of BPF prog after grace period") with regards to tail calls won't happen.
Integrating bpf_prog_free_deferred() directly into the RCU callback is
not allowed since the invocation might happen from either softirq or
process context, so we're not permitted to block. Reviewing all bpf_prog_put()
invocations from eBPF side (note, cBPF -> eBPF progs don't use this for
their destruction) with call_rcu() look good to me.
Since we don't know whether at the time of attaching the program, we're
already part of a tail call map, we need to use RCU variant. However, due
to this, there won't be severely more stress on the RCU callback queue:
situations with above bpf_prog_get() and bpf_prog_put() combo in practice
normally won't lead to releases, but even if they would, enough effort/
cycles have to be put into loading a BPF program into the kernel already.
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-30 23:24:43 +08:00
|
|
|
bpf_prog_put(ptr);
|
2015-08-06 15:02:33 +08:00
|
|
|
}
|
|
|
|
|
2017-06-28 14:08:34 +08:00
|
|
|
static u32 prog_fd_array_sys_lookup_elem(void *ptr)
|
|
|
|
{
|
|
|
|
return ((struct bpf_prog *)ptr)->aux->id;
|
|
|
|
}
|
|
|
|
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
/* decrement refcnt of all bpf_progs that are stored in this map */
|
2018-04-24 06:39:23 +08:00
|
|
|
static void bpf_fd_array_map_clear(struct bpf_map *map)
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < array->map.max_entries; i++)
|
2015-08-06 15:02:33 +08:00
|
|
|
fd_array_map_delete_elem(map, &i);
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
}
|
|
|
|
|
2017-04-11 21:34:58 +08:00
|
|
|
const struct bpf_map_ops prog_array_map_ops = {
|
2018-01-18 11:13:25 +08:00
|
|
|
.map_alloc_check = fd_array_map_alloc_check,
|
|
|
|
.map_alloc = array_map_alloc,
|
2015-08-06 15:02:33 +08:00
|
|
|
.map_free = fd_array_map_free,
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
.map_get_next_key = array_map_get_next_key,
|
2015-08-06 15:02:33 +08:00
|
|
|
.map_lookup_elem = fd_array_map_lookup_elem,
|
|
|
|
.map_delete_elem = fd_array_map_delete_elem,
|
|
|
|
.map_fd_get_ptr = prog_fd_array_get_ptr,
|
|
|
|
.map_fd_put_ptr = prog_fd_array_put_ptr,
|
2017-06-28 14:08:34 +08:00
|
|
|
.map_fd_sys_lookup_elem = prog_fd_array_sys_lookup_elem,
|
2018-04-24 06:39:23 +08:00
|
|
|
.map_release_uref = bpf_fd_array_map_clear,
|
bpf: allow bpf programs to tail-call other bpf programs
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-05-20 07:59:03 +08:00
|
|
|
};
|
|
|
|
|
bpf, maps: flush own entries on perf map release
The behavior of perf event arrays are quite different from all
others as they are tightly coupled to perf event fds, f.e. shown
recently by commit e03e7ee34fdd ("perf/bpf: Convert perf_event_array
to use struct file") to make refcounting on perf event more robust.
A remaining issue that the current code still has is that since
additions to the perf event array take a reference on the struct
file via perf_event_get() and are only released via fput() (that
cleans up the perf event eventually via perf_event_release_kernel())
when the element is either manually removed from the map from user
space or automatically when the last reference on the perf event
map is dropped. However, this leads us to dangling struct file's
when the map gets pinned after the application owning the perf
event descriptor exits, and since the struct file reference will
in such case only be manually dropped or via pinned file removal,
it leads to the perf event living longer than necessary, consuming
needlessly resources for that time.
Relations between perf event fds and bpf perf event map fds can be
rather complex. F.e. maps can act as demuxers among different perf
event fds that can possibly be owned by different threads and based
on the index selection from the program, events get dispatched to
one of the per-cpu fd endpoints. One perf event fd (or, rather a
per-cpu set of them) can also live in multiple perf event maps at
the same time, listening for events. Also, another requirement is
that perf event fds can get closed from application side after they
have been attached to the perf event map, so that on exit perf event
map will take care of dropping their references eventually. Likewise,
when such maps are pinned, the intended behavior is that a user
application does bpf_obj_get(), puts its fds in there and on exit
when fd is released, they are dropped from the map again, so the map
acts rather as connector endpoint. This also makes perf event maps
inherently different from program arrays as described in more detail
in commit c9da161c6517 ("bpf: fix clearing on persistent program
array maps").
To tackle this, map entries are marked by the map struct file that
added the element to the map. And when the last reference to that map
struct file is released from user space, then the tracked entries
are purged from the map. This is okay, because new map struct files
instances resp. frontends to the anon inode are provided via
bpf_map_new_fd() that is called when we invoke bpf_obj_get_user()
for retrieving a pinned map, but also when an initial instance is
created via map_create(). The rest is resolved by the vfs layer
automatically for us by keeping reference count on the map's struct
file. Any concurrent updates on the map slot are fine as well, it
just means that perf_event_fd_array_release() needs to delete less
of its own entires.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-16 04:47:14 +08:00
|
|
|
static struct bpf_event_entry *bpf_event_entry_gen(struct file *perf_file,
|
|
|
|
struct file *map_file)
|
2015-08-06 15:02:34 +08:00
|
|
|
{
|
bpf, maps: flush own entries on perf map release
The behavior of perf event arrays are quite different from all
others as they are tightly coupled to perf event fds, f.e. shown
recently by commit e03e7ee34fdd ("perf/bpf: Convert perf_event_array
to use struct file") to make refcounting on perf event more robust.
A remaining issue that the current code still has is that since
additions to the perf event array take a reference on the struct
file via perf_event_get() and are only released via fput() (that
cleans up the perf event eventually via perf_event_release_kernel())
when the element is either manually removed from the map from user
space or automatically when the last reference on the perf event
map is dropped. However, this leads us to dangling struct file's
when the map gets pinned after the application owning the perf
event descriptor exits, and since the struct file reference will
in such case only be manually dropped or via pinned file removal,
it leads to the perf event living longer than necessary, consuming
needlessly resources for that time.
Relations between perf event fds and bpf perf event map fds can be
rather complex. F.e. maps can act as demuxers among different perf
event fds that can possibly be owned by different threads and based
on the index selection from the program, events get dispatched to
one of the per-cpu fd endpoints. One perf event fd (or, rather a
per-cpu set of them) can also live in multiple perf event maps at
the same time, listening for events. Also, another requirement is
that perf event fds can get closed from application side after they
have been attached to the perf event map, so that on exit perf event
map will take care of dropping their references eventually. Likewise,
when such maps are pinned, the intended behavior is that a user
application does bpf_obj_get(), puts its fds in there and on exit
when fd is released, they are dropped from the map again, so the map
acts rather as connector endpoint. This also makes perf event maps
inherently different from program arrays as described in more detail
in commit c9da161c6517 ("bpf: fix clearing on persistent program
array maps").
To tackle this, map entries are marked by the map struct file that
added the element to the map. And when the last reference to that map
struct file is released from user space, then the tracked entries
are purged from the map. This is okay, because new map struct files
instances resp. frontends to the anon inode are provided via
bpf_map_new_fd() that is called when we invoke bpf_obj_get_user()
for retrieving a pinned map, but also when an initial instance is
created via map_create(). The rest is resolved by the vfs layer
automatically for us by keeping reference count on the map's struct
file. Any concurrent updates on the map slot are fine as well, it
just means that perf_event_fd_array_release() needs to delete less
of its own entires.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-16 04:47:14 +08:00
|
|
|
struct bpf_event_entry *ee;
|
|
|
|
|
2016-07-16 07:15:55 +08:00
|
|
|
ee = kzalloc(sizeof(*ee), GFP_ATOMIC);
|
bpf, maps: flush own entries on perf map release
The behavior of perf event arrays are quite different from all
others as they are tightly coupled to perf event fds, f.e. shown
recently by commit e03e7ee34fdd ("perf/bpf: Convert perf_event_array
to use struct file") to make refcounting on perf event more robust.
A remaining issue that the current code still has is that since
additions to the perf event array take a reference on the struct
file via perf_event_get() and are only released via fput() (that
cleans up the perf event eventually via perf_event_release_kernel())
when the element is either manually removed from the map from user
space or automatically when the last reference on the perf event
map is dropped. However, this leads us to dangling struct file's
when the map gets pinned after the application owning the perf
event descriptor exits, and since the struct file reference will
in such case only be manually dropped or via pinned file removal,
it leads to the perf event living longer than necessary, consuming
needlessly resources for that time.
Relations between perf event fds and bpf perf event map fds can be
rather complex. F.e. maps can act as demuxers among different perf
event fds that can possibly be owned by different threads and based
on the index selection from the program, events get dispatched to
one of the per-cpu fd endpoints. One perf event fd (or, rather a
per-cpu set of them) can also live in multiple perf event maps at
the same time, listening for events. Also, another requirement is
that perf event fds can get closed from application side after they
have been attached to the perf event map, so that on exit perf event
map will take care of dropping their references eventually. Likewise,
when such maps are pinned, the intended behavior is that a user
application does bpf_obj_get(), puts its fds in there and on exit
when fd is released, they are dropped from the map again, so the map
acts rather as connector endpoint. This also makes perf event maps
inherently different from program arrays as described in more detail
in commit c9da161c6517 ("bpf: fix clearing on persistent program
array maps").
To tackle this, map entries are marked by the map struct file that
added the element to the map. And when the last reference to that map
struct file is released from user space, then the tracked entries
are purged from the map. This is okay, because new map struct files
instances resp. frontends to the anon inode are provided via
bpf_map_new_fd() that is called when we invoke bpf_obj_get_user()
for retrieving a pinned map, but also when an initial instance is
created via map_create(). The rest is resolved by the vfs layer
automatically for us by keeping reference count on the map's struct
file. Any concurrent updates on the map slot are fine as well, it
just means that perf_event_fd_array_release() needs to delete less
of its own entires.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-16 04:47:14 +08:00
|
|
|
if (ee) {
|
|
|
|
ee->event = perf_file->private_data;
|
|
|
|
ee->perf_file = perf_file;
|
|
|
|
ee->map_file = map_file;
|
|
|
|
}
|
|
|
|
|
|
|
|
return ee;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __bpf_event_entry_free(struct rcu_head *rcu)
|
|
|
|
{
|
|
|
|
struct bpf_event_entry *ee;
|
|
|
|
|
|
|
|
ee = container_of(rcu, struct bpf_event_entry, rcu);
|
|
|
|
fput(ee->perf_file);
|
|
|
|
kfree(ee);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void bpf_event_entry_free_rcu(struct bpf_event_entry *ee)
|
|
|
|
{
|
|
|
|
call_rcu(&ee->rcu, __bpf_event_entry_free);
|
2015-08-06 15:02:34 +08:00
|
|
|
}
|
|
|
|
|
2016-06-16 04:47:13 +08:00
|
|
|
static void *perf_event_fd_array_get_ptr(struct bpf_map *map,
|
|
|
|
struct file *map_file, int fd)
|
2015-08-06 15:02:34 +08:00
|
|
|
{
|
bpf, maps: flush own entries on perf map release
The behavior of perf event arrays are quite different from all
others as they are tightly coupled to perf event fds, f.e. shown
recently by commit e03e7ee34fdd ("perf/bpf: Convert perf_event_array
to use struct file") to make refcounting on perf event more robust.
A remaining issue that the current code still has is that since
additions to the perf event array take a reference on the struct
file via perf_event_get() and are only released via fput() (that
cleans up the perf event eventually via perf_event_release_kernel())
when the element is either manually removed from the map from user
space or automatically when the last reference on the perf event
map is dropped. However, this leads us to dangling struct file's
when the map gets pinned after the application owning the perf
event descriptor exits, and since the struct file reference will
in such case only be manually dropped or via pinned file removal,
it leads to the perf event living longer than necessary, consuming
needlessly resources for that time.
Relations between perf event fds and bpf perf event map fds can be
rather complex. F.e. maps can act as demuxers among different perf
event fds that can possibly be owned by different threads and based
on the index selection from the program, events get dispatched to
one of the per-cpu fd endpoints. One perf event fd (or, rather a
per-cpu set of them) can also live in multiple perf event maps at
the same time, listening for events. Also, another requirement is
that perf event fds can get closed from application side after they
have been attached to the perf event map, so that on exit perf event
map will take care of dropping their references eventually. Likewise,
when such maps are pinned, the intended behavior is that a user
application does bpf_obj_get(), puts its fds in there and on exit
when fd is released, they are dropped from the map again, so the map
acts rather as connector endpoint. This also makes perf event maps
inherently different from program arrays as described in more detail
in commit c9da161c6517 ("bpf: fix clearing on persistent program
array maps").
To tackle this, map entries are marked by the map struct file that
added the element to the map. And when the last reference to that map
struct file is released from user space, then the tracked entries
are purged from the map. This is okay, because new map struct files
instances resp. frontends to the anon inode are provided via
bpf_map_new_fd() that is called when we invoke bpf_obj_get_user()
for retrieving a pinned map, but also when an initial instance is
created via map_create(). The rest is resolved by the vfs layer
automatically for us by keeping reference count on the map's struct
file. Any concurrent updates on the map slot are fine as well, it
just means that perf_event_fd_array_release() needs to delete less
of its own entires.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-16 04:47:14 +08:00
|
|
|
struct bpf_event_entry *ee;
|
|
|
|
struct perf_event *event;
|
|
|
|
struct file *perf_file;
|
2017-06-03 12:03:52 +08:00
|
|
|
u64 value;
|
2015-08-06 15:02:34 +08:00
|
|
|
|
bpf, maps: flush own entries on perf map release
The behavior of perf event arrays are quite different from all
others as they are tightly coupled to perf event fds, f.e. shown
recently by commit e03e7ee34fdd ("perf/bpf: Convert perf_event_array
to use struct file") to make refcounting on perf event more robust.
A remaining issue that the current code still has is that since
additions to the perf event array take a reference on the struct
file via perf_event_get() and are only released via fput() (that
cleans up the perf event eventually via perf_event_release_kernel())
when the element is either manually removed from the map from user
space or automatically when the last reference on the perf event
map is dropped. However, this leads us to dangling struct file's
when the map gets pinned after the application owning the perf
event descriptor exits, and since the struct file reference will
in such case only be manually dropped or via pinned file removal,
it leads to the perf event living longer than necessary, consuming
needlessly resources for that time.
Relations between perf event fds and bpf perf event map fds can be
rather complex. F.e. maps can act as demuxers among different perf
event fds that can possibly be owned by different threads and based
on the index selection from the program, events get dispatched to
one of the per-cpu fd endpoints. One perf event fd (or, rather a
per-cpu set of them) can also live in multiple perf event maps at
the same time, listening for events. Also, another requirement is
that perf event fds can get closed from application side after they
have been attached to the perf event map, so that on exit perf event
map will take care of dropping their references eventually. Likewise,
when such maps are pinned, the intended behavior is that a user
application does bpf_obj_get(), puts its fds in there and on exit
when fd is released, they are dropped from the map again, so the map
acts rather as connector endpoint. This also makes perf event maps
inherently different from program arrays as described in more detail
in commit c9da161c6517 ("bpf: fix clearing on persistent program
array maps").
To tackle this, map entries are marked by the map struct file that
added the element to the map. And when the last reference to that map
struct file is released from user space, then the tracked entries
are purged from the map. This is okay, because new map struct files
instances resp. frontends to the anon inode are provided via
bpf_map_new_fd() that is called when we invoke bpf_obj_get_user()
for retrieving a pinned map, but also when an initial instance is
created via map_create(). The rest is resolved by the vfs layer
automatically for us by keeping reference count on the map's struct
file. Any concurrent updates on the map slot are fine as well, it
just means that perf_event_fd_array_release() needs to delete less
of its own entires.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-16 04:47:14 +08:00
|
|
|
perf_file = perf_event_get(fd);
|
|
|
|
if (IS_ERR(perf_file))
|
|
|
|
return perf_file;
|
2016-01-26 12:59:49 +08:00
|
|
|
|
2017-06-03 12:03:52 +08:00
|
|
|
ee = ERR_PTR(-EOPNOTSUPP);
|
bpf, maps: flush own entries on perf map release
The behavior of perf event arrays are quite different from all
others as they are tightly coupled to perf event fds, f.e. shown
recently by commit e03e7ee34fdd ("perf/bpf: Convert perf_event_array
to use struct file") to make refcounting on perf event more robust.
A remaining issue that the current code still has is that since
additions to the perf event array take a reference on the struct
file via perf_event_get() and are only released via fput() (that
cleans up the perf event eventually via perf_event_release_kernel())
when the element is either manually removed from the map from user
space or automatically when the last reference on the perf event
map is dropped. However, this leads us to dangling struct file's
when the map gets pinned after the application owning the perf
event descriptor exits, and since the struct file reference will
in such case only be manually dropped or via pinned file removal,
it leads to the perf event living longer than necessary, consuming
needlessly resources for that time.
Relations between perf event fds and bpf perf event map fds can be
rather complex. F.e. maps can act as demuxers among different perf
event fds that can possibly be owned by different threads and based
on the index selection from the program, events get dispatched to
one of the per-cpu fd endpoints. One perf event fd (or, rather a
per-cpu set of them) can also live in multiple perf event maps at
the same time, listening for events. Also, another requirement is
that perf event fds can get closed from application side after they
have been attached to the perf event map, so that on exit perf event
map will take care of dropping their references eventually. Likewise,
when such maps are pinned, the intended behavior is that a user
application does bpf_obj_get(), puts its fds in there and on exit
when fd is released, they are dropped from the map again, so the map
acts rather as connector endpoint. This also makes perf event maps
inherently different from program arrays as described in more detail
in commit c9da161c6517 ("bpf: fix clearing on persistent program
array maps").
To tackle this, map entries are marked by the map struct file that
added the element to the map. And when the last reference to that map
struct file is released from user space, then the tracked entries
are purged from the map. This is okay, because new map struct files
instances resp. frontends to the anon inode are provided via
bpf_map_new_fd() that is called when we invoke bpf_obj_get_user()
for retrieving a pinned map, but also when an initial instance is
created via map_create(). The rest is resolved by the vfs layer
automatically for us by keeping reference count on the map's struct
file. Any concurrent updates on the map slot are fine as well, it
just means that perf_event_fd_array_release() needs to delete less
of its own entires.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-16 04:47:14 +08:00
|
|
|
event = perf_file->private_data;
|
2017-10-06 00:19:19 +08:00
|
|
|
if (perf_event_read_local(event, &value, NULL, NULL) == -EOPNOTSUPP)
|
bpf, maps: flush own entries on perf map release
The behavior of perf event arrays are quite different from all
others as they are tightly coupled to perf event fds, f.e. shown
recently by commit e03e7ee34fdd ("perf/bpf: Convert perf_event_array
to use struct file") to make refcounting on perf event more robust.
A remaining issue that the current code still has is that since
additions to the perf event array take a reference on the struct
file via perf_event_get() and are only released via fput() (that
cleans up the perf event eventually via perf_event_release_kernel())
when the element is either manually removed from the map from user
space or automatically when the last reference on the perf event
map is dropped. However, this leads us to dangling struct file's
when the map gets pinned after the application owning the perf
event descriptor exits, and since the struct file reference will
in such case only be manually dropped or via pinned file removal,
it leads to the perf event living longer than necessary, consuming
needlessly resources for that time.
Relations between perf event fds and bpf perf event map fds can be
rather complex. F.e. maps can act as demuxers among different perf
event fds that can possibly be owned by different threads and based
on the index selection from the program, events get dispatched to
one of the per-cpu fd endpoints. One perf event fd (or, rather a
per-cpu set of them) can also live in multiple perf event maps at
the same time, listening for events. Also, another requirement is
that perf event fds can get closed from application side after they
have been attached to the perf event map, so that on exit perf event
map will take care of dropping their references eventually. Likewise,
when such maps are pinned, the intended behavior is that a user
application does bpf_obj_get(), puts its fds in there and on exit
when fd is released, they are dropped from the map again, so the map
acts rather as connector endpoint. This also makes perf event maps
inherently different from program arrays as described in more detail
in commit c9da161c6517 ("bpf: fix clearing on persistent program
array maps").
To tackle this, map entries are marked by the map struct file that
added the element to the map. And when the last reference to that map
struct file is released from user space, then the tracked entries
are purged from the map. This is okay, because new map struct files
instances resp. frontends to the anon inode are provided via
bpf_map_new_fd() that is called when we invoke bpf_obj_get_user()
for retrieving a pinned map, but also when an initial instance is
created via map_create(). The rest is resolved by the vfs layer
automatically for us by keeping reference count on the map's struct
file. Any concurrent updates on the map slot are fine as well, it
just means that perf_event_fd_array_release() needs to delete less
of its own entires.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-16 04:47:14 +08:00
|
|
|
goto err_out;
|
|
|
|
|
2017-06-03 12:03:52 +08:00
|
|
|
ee = bpf_event_entry_gen(perf_file, map_file);
|
|
|
|
if (ee)
|
|
|
|
return ee;
|
|
|
|
ee = ERR_PTR(-ENOMEM);
|
bpf, maps: flush own entries on perf map release
The behavior of perf event arrays are quite different from all
others as they are tightly coupled to perf event fds, f.e. shown
recently by commit e03e7ee34fdd ("perf/bpf: Convert perf_event_array
to use struct file") to make refcounting on perf event more robust.
A remaining issue that the current code still has is that since
additions to the perf event array take a reference on the struct
file via perf_event_get() and are only released via fput() (that
cleans up the perf event eventually via perf_event_release_kernel())
when the element is either manually removed from the map from user
space or automatically when the last reference on the perf event
map is dropped. However, this leads us to dangling struct file's
when the map gets pinned after the application owning the perf
event descriptor exits, and since the struct file reference will
in such case only be manually dropped or via pinned file removal,
it leads to the perf event living longer than necessary, consuming
needlessly resources for that time.
Relations between perf event fds and bpf perf event map fds can be
rather complex. F.e. maps can act as demuxers among different perf
event fds that can possibly be owned by different threads and based
on the index selection from the program, events get dispatched to
one of the per-cpu fd endpoints. One perf event fd (or, rather a
per-cpu set of them) can also live in multiple perf event maps at
the same time, listening for events. Also, another requirement is
that perf event fds can get closed from application side after they
have been attached to the perf event map, so that on exit perf event
map will take care of dropping their references eventually. Likewise,
when such maps are pinned, the intended behavior is that a user
application does bpf_obj_get(), puts its fds in there and on exit
when fd is released, they are dropped from the map again, so the map
acts rather as connector endpoint. This also makes perf event maps
inherently different from program arrays as described in more detail
in commit c9da161c6517 ("bpf: fix clearing on persistent program
array maps").
To tackle this, map entries are marked by the map struct file that
added the element to the map. And when the last reference to that map
struct file is released from user space, then the tracked entries
are purged from the map. This is okay, because new map struct files
instances resp. frontends to the anon inode are provided via
bpf_map_new_fd() that is called when we invoke bpf_obj_get_user()
for retrieving a pinned map, but also when an initial instance is
created via map_create(). The rest is resolved by the vfs layer
automatically for us by keeping reference count on the map's struct
file. Any concurrent updates on the map slot are fine as well, it
just means that perf_event_fd_array_release() needs to delete less
of its own entires.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-16 04:47:14 +08:00
|
|
|
err_out:
|
|
|
|
fput(perf_file);
|
|
|
|
return ee;
|
2015-08-06 15:02:34 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static void perf_event_fd_array_put_ptr(void *ptr)
|
|
|
|
{
|
bpf, maps: flush own entries on perf map release
The behavior of perf event arrays are quite different from all
others as they are tightly coupled to perf event fds, f.e. shown
recently by commit e03e7ee34fdd ("perf/bpf: Convert perf_event_array
to use struct file") to make refcounting on perf event more robust.
A remaining issue that the current code still has is that since
additions to the perf event array take a reference on the struct
file via perf_event_get() and are only released via fput() (that
cleans up the perf event eventually via perf_event_release_kernel())
when the element is either manually removed from the map from user
space or automatically when the last reference on the perf event
map is dropped. However, this leads us to dangling struct file's
when the map gets pinned after the application owning the perf
event descriptor exits, and since the struct file reference will
in such case only be manually dropped or via pinned file removal,
it leads to the perf event living longer than necessary, consuming
needlessly resources for that time.
Relations between perf event fds and bpf perf event map fds can be
rather complex. F.e. maps can act as demuxers among different perf
event fds that can possibly be owned by different threads and based
on the index selection from the program, events get dispatched to
one of the per-cpu fd endpoints. One perf event fd (or, rather a
per-cpu set of them) can also live in multiple perf event maps at
the same time, listening for events. Also, another requirement is
that perf event fds can get closed from application side after they
have been attached to the perf event map, so that on exit perf event
map will take care of dropping their references eventually. Likewise,
when such maps are pinned, the intended behavior is that a user
application does bpf_obj_get(), puts its fds in there and on exit
when fd is released, they are dropped from the map again, so the map
acts rather as connector endpoint. This also makes perf event maps
inherently different from program arrays as described in more detail
in commit c9da161c6517 ("bpf: fix clearing on persistent program
array maps").
To tackle this, map entries are marked by the map struct file that
added the element to the map. And when the last reference to that map
struct file is released from user space, then the tracked entries
are purged from the map. This is okay, because new map struct files
instances resp. frontends to the anon inode are provided via
bpf_map_new_fd() that is called when we invoke bpf_obj_get_user()
for retrieving a pinned map, but also when an initial instance is
created via map_create(). The rest is resolved by the vfs layer
automatically for us by keeping reference count on the map's struct
file. Any concurrent updates on the map slot are fine as well, it
just means that perf_event_fd_array_release() needs to delete less
of its own entires.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-16 04:47:14 +08:00
|
|
|
bpf_event_entry_free_rcu(ptr);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void perf_event_fd_array_release(struct bpf_map *map,
|
|
|
|
struct file *map_file)
|
|
|
|
{
|
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
|
|
|
struct bpf_event_entry *ee;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
for (i = 0; i < array->map.max_entries; i++) {
|
|
|
|
ee = READ_ONCE(array->ptrs[i]);
|
|
|
|
if (ee && ee->map_file == map_file)
|
|
|
|
fd_array_map_delete_elem(map, &i);
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
2015-08-06 15:02:34 +08:00
|
|
|
}
|
|
|
|
|
2017-04-11 21:34:58 +08:00
|
|
|
const struct bpf_map_ops perf_event_array_map_ops = {
|
2018-01-18 11:13:25 +08:00
|
|
|
.map_alloc_check = fd_array_map_alloc_check,
|
|
|
|
.map_alloc = array_map_alloc,
|
bpf, maps: flush own entries on perf map release
The behavior of perf event arrays are quite different from all
others as they are tightly coupled to perf event fds, f.e. shown
recently by commit e03e7ee34fdd ("perf/bpf: Convert perf_event_array
to use struct file") to make refcounting on perf event more robust.
A remaining issue that the current code still has is that since
additions to the perf event array take a reference on the struct
file via perf_event_get() and are only released via fput() (that
cleans up the perf event eventually via perf_event_release_kernel())
when the element is either manually removed from the map from user
space or automatically when the last reference on the perf event
map is dropped. However, this leads us to dangling struct file's
when the map gets pinned after the application owning the perf
event descriptor exits, and since the struct file reference will
in such case only be manually dropped or via pinned file removal,
it leads to the perf event living longer than necessary, consuming
needlessly resources for that time.
Relations between perf event fds and bpf perf event map fds can be
rather complex. F.e. maps can act as demuxers among different perf
event fds that can possibly be owned by different threads and based
on the index selection from the program, events get dispatched to
one of the per-cpu fd endpoints. One perf event fd (or, rather a
per-cpu set of them) can also live in multiple perf event maps at
the same time, listening for events. Also, another requirement is
that perf event fds can get closed from application side after they
have been attached to the perf event map, so that on exit perf event
map will take care of dropping their references eventually. Likewise,
when such maps are pinned, the intended behavior is that a user
application does bpf_obj_get(), puts its fds in there and on exit
when fd is released, they are dropped from the map again, so the map
acts rather as connector endpoint. This also makes perf event maps
inherently different from program arrays as described in more detail
in commit c9da161c6517 ("bpf: fix clearing on persistent program
array maps").
To tackle this, map entries are marked by the map struct file that
added the element to the map. And when the last reference to that map
struct file is released from user space, then the tracked entries
are purged from the map. This is okay, because new map struct files
instances resp. frontends to the anon inode are provided via
bpf_map_new_fd() that is called when we invoke bpf_obj_get_user()
for retrieving a pinned map, but also when an initial instance is
created via map_create(). The rest is resolved by the vfs layer
automatically for us by keeping reference count on the map's struct
file. Any concurrent updates on the map slot are fine as well, it
just means that perf_event_fd_array_release() needs to delete less
of its own entires.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-16 04:47:14 +08:00
|
|
|
.map_free = fd_array_map_free,
|
2015-08-06 15:02:34 +08:00
|
|
|
.map_get_next_key = array_map_get_next_key,
|
|
|
|
.map_lookup_elem = fd_array_map_lookup_elem,
|
|
|
|
.map_delete_elem = fd_array_map_delete_elem,
|
|
|
|
.map_fd_get_ptr = perf_event_fd_array_get_ptr,
|
|
|
|
.map_fd_put_ptr = perf_event_fd_array_put_ptr,
|
bpf, maps: flush own entries on perf map release
The behavior of perf event arrays are quite different from all
others as they are tightly coupled to perf event fds, f.e. shown
recently by commit e03e7ee34fdd ("perf/bpf: Convert perf_event_array
to use struct file") to make refcounting on perf event more robust.
A remaining issue that the current code still has is that since
additions to the perf event array take a reference on the struct
file via perf_event_get() and are only released via fput() (that
cleans up the perf event eventually via perf_event_release_kernel())
when the element is either manually removed from the map from user
space or automatically when the last reference on the perf event
map is dropped. However, this leads us to dangling struct file's
when the map gets pinned after the application owning the perf
event descriptor exits, and since the struct file reference will
in such case only be manually dropped or via pinned file removal,
it leads to the perf event living longer than necessary, consuming
needlessly resources for that time.
Relations between perf event fds and bpf perf event map fds can be
rather complex. F.e. maps can act as demuxers among different perf
event fds that can possibly be owned by different threads and based
on the index selection from the program, events get dispatched to
one of the per-cpu fd endpoints. One perf event fd (or, rather a
per-cpu set of them) can also live in multiple perf event maps at
the same time, listening for events. Also, another requirement is
that perf event fds can get closed from application side after they
have been attached to the perf event map, so that on exit perf event
map will take care of dropping their references eventually. Likewise,
when such maps are pinned, the intended behavior is that a user
application does bpf_obj_get(), puts its fds in there and on exit
when fd is released, they are dropped from the map again, so the map
acts rather as connector endpoint. This also makes perf event maps
inherently different from program arrays as described in more detail
in commit c9da161c6517 ("bpf: fix clearing on persistent program
array maps").
To tackle this, map entries are marked by the map struct file that
added the element to the map. And when the last reference to that map
struct file is released from user space, then the tracked entries
are purged from the map. This is okay, because new map struct files
instances resp. frontends to the anon inode are provided via
bpf_map_new_fd() that is called when we invoke bpf_obj_get_user()
for retrieving a pinned map, but also when an initial instance is
created via map_create(). The rest is resolved by the vfs layer
automatically for us by keeping reference count on the map's struct
file. Any concurrent updates on the map slot are fine as well, it
just means that perf_event_fd_array_release() needs to delete less
of its own entires.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-06-16 04:47:14 +08:00
|
|
|
.map_release = perf_event_fd_array_release,
|
2015-08-06 15:02:34 +08:00
|
|
|
};
|
|
|
|
|
2016-08-12 23:56:52 +08:00
|
|
|
#ifdef CONFIG_CGROUPS
|
2016-07-01 01:28:43 +08:00
|
|
|
static void *cgroup_fd_array_get_ptr(struct bpf_map *map,
|
|
|
|
struct file *map_file /* not used */,
|
|
|
|
int fd)
|
|
|
|
{
|
|
|
|
return cgroup_get_from_fd(fd);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void cgroup_fd_array_put_ptr(void *ptr)
|
|
|
|
{
|
|
|
|
/* cgroup_put free cgrp after a rcu grace period */
|
|
|
|
cgroup_put(ptr);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void cgroup_fd_array_free(struct bpf_map *map)
|
|
|
|
{
|
|
|
|
bpf_fd_array_map_clear(map);
|
|
|
|
fd_array_map_free(map);
|
|
|
|
}
|
|
|
|
|
2017-04-11 21:34:58 +08:00
|
|
|
const struct bpf_map_ops cgroup_array_map_ops = {
|
2018-01-18 11:13:25 +08:00
|
|
|
.map_alloc_check = fd_array_map_alloc_check,
|
|
|
|
.map_alloc = array_map_alloc,
|
2016-07-01 01:28:43 +08:00
|
|
|
.map_free = cgroup_fd_array_free,
|
|
|
|
.map_get_next_key = array_map_get_next_key,
|
|
|
|
.map_lookup_elem = fd_array_map_lookup_elem,
|
|
|
|
.map_delete_elem = fd_array_map_delete_elem,
|
|
|
|
.map_fd_get_ptr = cgroup_fd_array_get_ptr,
|
|
|
|
.map_fd_put_ptr = cgroup_fd_array_put_ptr,
|
|
|
|
};
|
|
|
|
#endif
|
2017-03-23 01:00:33 +08:00
|
|
|
|
|
|
|
static struct bpf_map *array_of_map_alloc(union bpf_attr *attr)
|
|
|
|
{
|
|
|
|
struct bpf_map *map, *inner_map_meta;
|
|
|
|
|
|
|
|
inner_map_meta = bpf_map_meta_alloc(attr->inner_map_fd);
|
|
|
|
if (IS_ERR(inner_map_meta))
|
|
|
|
return inner_map_meta;
|
|
|
|
|
2018-01-18 11:13:25 +08:00
|
|
|
map = array_map_alloc(attr);
|
2017-03-23 01:00:33 +08:00
|
|
|
if (IS_ERR(map)) {
|
|
|
|
bpf_map_meta_free(inner_map_meta);
|
|
|
|
return map;
|
|
|
|
}
|
|
|
|
|
|
|
|
map->inner_map_meta = inner_map_meta;
|
|
|
|
|
|
|
|
return map;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void array_of_map_free(struct bpf_map *map)
|
|
|
|
{
|
|
|
|
/* map->inner_map_meta is only accessed by syscall which
|
|
|
|
* is protected by fdget/fdput.
|
|
|
|
*/
|
|
|
|
bpf_map_meta_free(map->inner_map_meta);
|
|
|
|
bpf_fd_array_map_clear(map);
|
|
|
|
fd_array_map_free(map);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void *array_of_map_lookup_elem(struct bpf_map *map, void *key)
|
|
|
|
{
|
|
|
|
struct bpf_map **inner_map = array_map_lookup_elem(map, key);
|
|
|
|
|
|
|
|
if (!inner_map)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
return READ_ONCE(*inner_map);
|
|
|
|
}
|
|
|
|
|
2017-08-19 09:12:46 +08:00
|
|
|
static u32 array_of_map_gen_lookup(struct bpf_map *map,
|
|
|
|
struct bpf_insn *insn_buf)
|
|
|
|
{
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
struct bpf_array *array = container_of(map, struct bpf_array, map);
|
2017-08-19 09:12:46 +08:00
|
|
|
u32 elem_size = round_up(map->value_size, 8);
|
|
|
|
struct bpf_insn *insn = insn_buf;
|
|
|
|
const int ret = BPF_REG_0;
|
|
|
|
const int map_ptr = BPF_REG_1;
|
|
|
|
const int index = BPF_REG_2;
|
|
|
|
|
|
|
|
*insn++ = BPF_ALU64_IMM(BPF_ADD, map_ptr, offsetof(struct bpf_array, value));
|
|
|
|
*insn++ = BPF_LDX_MEM(BPF_W, ret, index, 0);
|
bpf: prevent out-of-bounds speculation
Under speculation, CPUs may mis-predict branches in bounds checks. Thus,
memory accesses under a bounds check may be speculated even if the
bounds check fails, providing a primitive for building a side channel.
To avoid leaking kernel data round up array-based maps and mask the index
after bounds check, so speculated load with out of bounds index will load
either valid value from the array or zero from the padded area.
Unconditionally mask index for all array types even when max_entries
are not rounded to power of 2 for root user.
When map is created by unpriv user generate a sequence of bpf insns
that includes AND operation to make sure that JITed code includes
the same 'index & index_mask' operation.
If prog_array map is created by unpriv user replace
bpf_tail_call(ctx, map, index);
with
if (index >= max_entries) {
index &= map->index_mask;
bpf_tail_call(ctx, map, index);
}
(along with roundup to power 2) to prevent out-of-bounds speculation.
There is secondary redundant 'if (index >= max_entries)' in the interpreter
and in all JITs, but they can be optimized later if necessary.
Other array-like maps (cpumap, devmap, sockmap, perf_event_array, cgroup_array)
cannot be used by unpriv, so no changes there.
That fixes bpf side of "Variant 1: bounds check bypass (CVE-2017-5753)" on
all architectures with and without JIT.
v2->v3:
Daniel noticed that attack potentially can be crafted via syscall commands
without loading the program, so add masking to those paths as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
2018-01-08 09:33:02 +08:00
|
|
|
if (map->unpriv_array) {
|
|
|
|
*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 6);
|
|
|
|
*insn++ = BPF_ALU32_IMM(BPF_AND, ret, array->index_mask);
|
|
|
|
} else {
|
|
|
|
*insn++ = BPF_JMP_IMM(BPF_JGE, ret, map->max_entries, 5);
|
|
|
|
}
|
2017-08-19 09:12:46 +08:00
|
|
|
if (is_power_of_2(elem_size))
|
|
|
|
*insn++ = BPF_ALU64_IMM(BPF_LSH, ret, ilog2(elem_size));
|
|
|
|
else
|
|
|
|
*insn++ = BPF_ALU64_IMM(BPF_MUL, ret, elem_size);
|
|
|
|
*insn++ = BPF_ALU64_REG(BPF_ADD, ret, map_ptr);
|
|
|
|
*insn++ = BPF_LDX_MEM(BPF_DW, ret, ret, 0);
|
|
|
|
*insn++ = BPF_JMP_IMM(BPF_JEQ, ret, 0, 1);
|
|
|
|
*insn++ = BPF_JMP_IMM(BPF_JA, 0, 0, 1);
|
|
|
|
*insn++ = BPF_MOV64_IMM(ret, 0);
|
|
|
|
|
|
|
|
return insn - insn_buf;
|
|
|
|
}
|
|
|
|
|
2017-04-11 21:34:58 +08:00
|
|
|
const struct bpf_map_ops array_of_maps_map_ops = {
|
2018-01-18 11:13:25 +08:00
|
|
|
.map_alloc_check = fd_array_map_alloc_check,
|
2017-03-23 01:00:33 +08:00
|
|
|
.map_alloc = array_of_map_alloc,
|
|
|
|
.map_free = array_of_map_free,
|
|
|
|
.map_get_next_key = array_map_get_next_key,
|
|
|
|
.map_lookup_elem = array_of_map_lookup_elem,
|
|
|
|
.map_delete_elem = fd_array_map_delete_elem,
|
|
|
|
.map_fd_get_ptr = bpf_map_fd_get_ptr,
|
|
|
|
.map_fd_put_ptr = bpf_map_fd_put_ptr,
|
2017-06-28 14:08:34 +08:00
|
|
|
.map_fd_sys_lookup_elem = bpf_map_fd_sys_lookup_elem,
|
2017-08-19 09:12:46 +08:00
|
|
|
.map_gen_lookup = array_of_map_gen_lookup,
|
2017-03-23 01:00:33 +08:00
|
|
|
};
|