linux/net/sched/act_bpf.c

441 lines
9.9 KiB
C
Raw Normal View History

/*
* Copyright (c) 2015 Jiri Pirko <jiri@resnulli.us>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*/
#include <linux/module.h>
#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/skbuff.h>
#include <linux/rtnetlink.h>
#include <linux/filter.h>
#include <linux/bpf.h>
#include <net/netlink.h>
#include <net/pkt_sched.h>
#include <linux/tc_act/tc_bpf.h>
#include <net/tc_act/tc_bpf.h>
#define BPF_TAB_MASK 15
#define ACT_BPF_NAME_LEN 256
struct tcf_bpf_cfg {
struct bpf_prog *filter;
struct sock_filter *bpf_ops;
act_bpf: fix memory leaks when replacing bpf programs We currently trigger multiple memory leaks when replacing bpf actions, besides others: comm "tc", pid 1909, jiffies 4294851310 (age 1602.796s) hex dump (first 32 bytes): 01 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 ................ 18 b0 98 6d 00 88 ff ff 00 00 00 00 00 00 00 00 ...m............ backtrace: [<ffffffff817e623e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff8120a22d>] __vmalloc_node_range+0x1bd/0x2c0 [<ffffffff8120a37a>] __vmalloc+0x4a/0x50 [<ffffffff811a8d0a>] bpf_prog_alloc+0x3a/0xa0 [<ffffffff816c0684>] bpf_prog_create+0x44/0xa0 [<ffffffffa09ba4eb>] tcf_bpf_init+0x28b/0x3c0 [act_bpf] [<ffffffff816d7001>] tcf_action_init_1+0x191/0x1b0 [<ffffffff816d70a2>] tcf_action_init+0x82/0xf0 [<ffffffff816d4d12>] tcf_exts_validate+0xb2/0xc0 [<ffffffffa09b5838>] cls_bpf_modify_existing+0x98/0x340 [cls_bpf] [<ffffffffa09b5cd6>] cls_bpf_change+0x1a6/0x274 [cls_bpf] [<ffffffff816d56e5>] tc_ctl_tfilter+0x335/0x910 [<ffffffff816b9145>] rtnetlink_rcv_msg+0x95/0x240 [<ffffffff816df34f>] netlink_rcv_skb+0xaf/0xc0 [<ffffffff816b909e>] rtnetlink_rcv+0x2e/0x40 [<ffffffff816deaaf>] netlink_unicast+0xef/0x1b0 Issue is that the old content from tcf_bpf is allocated and needs to be released when we replace it. We seem to do that since the beginning of act_bpf on the filter and insns, later on the name as well. Example test case, after patch: # FOO="1,6 0 0 4294967295," # BAR="1,6 0 0 4294967294," # tc actions add action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$BAR" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967294' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions del action bpf index 2 [...] # echo "scan" > /sys/kernel/debug/kmemleak # cat /sys/kernel/debug/kmemleak | grep "comm \"tc\"" | wc -l 0 Fixes: d23b8ad8ab23 ("tc: add BPF based action") Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-07-29 16:40:56 +00:00
const char *bpf_name;
u32 bpf_fd;
u16 bpf_num_ops;
act_bpf: fix memory leaks when replacing bpf programs We currently trigger multiple memory leaks when replacing bpf actions, besides others: comm "tc", pid 1909, jiffies 4294851310 (age 1602.796s) hex dump (first 32 bytes): 01 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 ................ 18 b0 98 6d 00 88 ff ff 00 00 00 00 00 00 00 00 ...m............ backtrace: [<ffffffff817e623e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff8120a22d>] __vmalloc_node_range+0x1bd/0x2c0 [<ffffffff8120a37a>] __vmalloc+0x4a/0x50 [<ffffffff811a8d0a>] bpf_prog_alloc+0x3a/0xa0 [<ffffffff816c0684>] bpf_prog_create+0x44/0xa0 [<ffffffffa09ba4eb>] tcf_bpf_init+0x28b/0x3c0 [act_bpf] [<ffffffff816d7001>] tcf_action_init_1+0x191/0x1b0 [<ffffffff816d70a2>] tcf_action_init+0x82/0xf0 [<ffffffff816d4d12>] tcf_exts_validate+0xb2/0xc0 [<ffffffffa09b5838>] cls_bpf_modify_existing+0x98/0x340 [cls_bpf] [<ffffffffa09b5cd6>] cls_bpf_change+0x1a6/0x274 [cls_bpf] [<ffffffff816d56e5>] tc_ctl_tfilter+0x335/0x910 [<ffffffff816b9145>] rtnetlink_rcv_msg+0x95/0x240 [<ffffffff816df34f>] netlink_rcv_skb+0xaf/0xc0 [<ffffffff816b909e>] rtnetlink_rcv+0x2e/0x40 [<ffffffff816deaaf>] netlink_unicast+0xef/0x1b0 Issue is that the old content from tcf_bpf is allocated and needs to be released when we replace it. We seem to do that since the beginning of act_bpf on the filter and insns, later on the name as well. Example test case, after patch: # FOO="1,6 0 0 4294967295," # BAR="1,6 0 0 4294967294," # tc actions add action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$BAR" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967294' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions del action bpf index 2 [...] # echo "scan" > /sys/kernel/debug/kmemleak # cat /sys/kernel/debug/kmemleak | grep "comm \"tc\"" | wc -l 0 Fixes: d23b8ad8ab23 ("tc: add BPF based action") Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-07-29 16:40:56 +00:00
bool is_ebpf;
};
static int bpf_net_id;
static int tcf_bpf(struct sk_buff *skb, const struct tc_action *act,
struct tcf_result *res)
{
struct tcf_bpf *prog = act->priv;
struct bpf_prog *filter;
act_bpf: allow non-default TC_ACT opcodes as BPF exec outcome Revisiting commit d23b8ad8ab23 ("tc: add BPF based action") with regards to eBPF support, I was thinking that it might be better to improve return semantics from a BPF program invoked through BPF_PROG_RUN(). Currently, in case filter_res is 0, we overwrite the default action opcode with TC_ACT_SHOT. A default action opcode configured through tc's m_bpf can be: TC_ACT_RECLASSIFY, TC_ACT_PIPE, TC_ACT_SHOT, TC_ACT_UNSPEC, TC_ACT_OK. In cls_bpf, we have the possibility to overwrite the default class associated with the classifier in case filter_res is _not_ 0xffffffff (-1). That allows us to fold multiple [e]BPF programs into a single one, where they would otherwise need to be defined as a separate classifier with its own classid, needlessly redoing parsing work, etc. Similarly, we could do better in act_bpf: Since above TC_ACT* opcodes are exported to UAPI anyway, we reuse them for return-code-to-tc-opcode mapping, where we would allow above possibilities. Thus, like in cls_bpf, a filter_res of 0xffffffff (-1) means that the configured _default_ action is used. Any unkown return code from the BPF program would fail in tcf_bpf() with TC_ACT_UNSPEC. Should we one day want to make use of TC_ACT_STOLEN or TC_ACT_QUEUED, which both have the same semantics, we have the option to either use that as a default action (filter_res of 0xffffffff) or non-default BPF return code. All that will allow us to transparently use tcf_bpf() for both BPF flavours. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Cc: Jiri Pirko <jiri@resnulli.us> Cc: Alexei Starovoitov <ast@plumgrid.com> Cc: Jamal Hadi Salim <jhs@mojatatu.com> Acked-by: Jiri Pirko <jiri@resnulli.us> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-03-17 19:25:57 +00:00
int action, filter_res;
bool at_ingress = G_TC_AT(skb->tc_verd) & AT_INGRESS;
if (unlikely(!skb_mac_header_was_set(skb)))
return TC_ACT_UNSPEC;
tcf_lastuse_update(&prog->tcf_tm);
bstats_cpu_update(this_cpu_ptr(prog->common.cpu_bstats), skb);
rcu_read_lock();
filter = rcu_dereference(prog->filter);
if (at_ingress) {
__skb_push(skb, skb->mac_len);
bpf_compute_data_end(skb);
filter_res = BPF_PROG_RUN(filter, skb);
__skb_pull(skb, skb->mac_len);
} else {
bpf_compute_data_end(skb);
filter_res = BPF_PROG_RUN(filter, skb);
}
rcu_read_unlock();
act_bpf: allow non-default TC_ACT opcodes as BPF exec outcome Revisiting commit d23b8ad8ab23 ("tc: add BPF based action") with regards to eBPF support, I was thinking that it might be better to improve return semantics from a BPF program invoked through BPF_PROG_RUN(). Currently, in case filter_res is 0, we overwrite the default action opcode with TC_ACT_SHOT. A default action opcode configured through tc's m_bpf can be: TC_ACT_RECLASSIFY, TC_ACT_PIPE, TC_ACT_SHOT, TC_ACT_UNSPEC, TC_ACT_OK. In cls_bpf, we have the possibility to overwrite the default class associated with the classifier in case filter_res is _not_ 0xffffffff (-1). That allows us to fold multiple [e]BPF programs into a single one, where they would otherwise need to be defined as a separate classifier with its own classid, needlessly redoing parsing work, etc. Similarly, we could do better in act_bpf: Since above TC_ACT* opcodes are exported to UAPI anyway, we reuse them for return-code-to-tc-opcode mapping, where we would allow above possibilities. Thus, like in cls_bpf, a filter_res of 0xffffffff (-1) means that the configured _default_ action is used. Any unkown return code from the BPF program would fail in tcf_bpf() with TC_ACT_UNSPEC. Should we one day want to make use of TC_ACT_STOLEN or TC_ACT_QUEUED, which both have the same semantics, we have the option to either use that as a default action (filter_res of 0xffffffff) or non-default BPF return code. All that will allow us to transparently use tcf_bpf() for both BPF flavours. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Cc: Jiri Pirko <jiri@resnulli.us> Cc: Alexei Starovoitov <ast@plumgrid.com> Cc: Jamal Hadi Salim <jhs@mojatatu.com> Acked-by: Jiri Pirko <jiri@resnulli.us> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-03-17 19:25:57 +00:00
/* A BPF program may overwrite the default action opcode.
* Similarly as in cls_bpf, if filter_res == -1 we use the
* default action specified from tc.
*
* In case a different well-known TC_ACT opcode has been
* returned, it will overwrite the default one.
*
* For everything else that is unkown, TC_ACT_UNSPEC is
* returned.
*/
switch (filter_res) {
case TC_ACT_PIPE:
case TC_ACT_RECLASSIFY:
case TC_ACT_OK:
2015-09-16 06:05:43 +00:00
case TC_ACT_REDIRECT:
act_bpf: allow non-default TC_ACT opcodes as BPF exec outcome Revisiting commit d23b8ad8ab23 ("tc: add BPF based action") with regards to eBPF support, I was thinking that it might be better to improve return semantics from a BPF program invoked through BPF_PROG_RUN(). Currently, in case filter_res is 0, we overwrite the default action opcode with TC_ACT_SHOT. A default action opcode configured through tc's m_bpf can be: TC_ACT_RECLASSIFY, TC_ACT_PIPE, TC_ACT_SHOT, TC_ACT_UNSPEC, TC_ACT_OK. In cls_bpf, we have the possibility to overwrite the default class associated with the classifier in case filter_res is _not_ 0xffffffff (-1). That allows us to fold multiple [e]BPF programs into a single one, where they would otherwise need to be defined as a separate classifier with its own classid, needlessly redoing parsing work, etc. Similarly, we could do better in act_bpf: Since above TC_ACT* opcodes are exported to UAPI anyway, we reuse them for return-code-to-tc-opcode mapping, where we would allow above possibilities. Thus, like in cls_bpf, a filter_res of 0xffffffff (-1) means that the configured _default_ action is used. Any unkown return code from the BPF program would fail in tcf_bpf() with TC_ACT_UNSPEC. Should we one day want to make use of TC_ACT_STOLEN or TC_ACT_QUEUED, which both have the same semantics, we have the option to either use that as a default action (filter_res of 0xffffffff) or non-default BPF return code. All that will allow us to transparently use tcf_bpf() for both BPF flavours. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Cc: Jiri Pirko <jiri@resnulli.us> Cc: Alexei Starovoitov <ast@plumgrid.com> Cc: Jamal Hadi Salim <jhs@mojatatu.com> Acked-by: Jiri Pirko <jiri@resnulli.us> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-03-17 19:25:57 +00:00
action = filter_res;
break;
case TC_ACT_SHOT:
action = filter_res;
qstats_drop_inc(this_cpu_ptr(prog->common.cpu_qstats));
act_bpf: allow non-default TC_ACT opcodes as BPF exec outcome Revisiting commit d23b8ad8ab23 ("tc: add BPF based action") with regards to eBPF support, I was thinking that it might be better to improve return semantics from a BPF program invoked through BPF_PROG_RUN(). Currently, in case filter_res is 0, we overwrite the default action opcode with TC_ACT_SHOT. A default action opcode configured through tc's m_bpf can be: TC_ACT_RECLASSIFY, TC_ACT_PIPE, TC_ACT_SHOT, TC_ACT_UNSPEC, TC_ACT_OK. In cls_bpf, we have the possibility to overwrite the default class associated with the classifier in case filter_res is _not_ 0xffffffff (-1). That allows us to fold multiple [e]BPF programs into a single one, where they would otherwise need to be defined as a separate classifier with its own classid, needlessly redoing parsing work, etc. Similarly, we could do better in act_bpf: Since above TC_ACT* opcodes are exported to UAPI anyway, we reuse them for return-code-to-tc-opcode mapping, where we would allow above possibilities. Thus, like in cls_bpf, a filter_res of 0xffffffff (-1) means that the configured _default_ action is used. Any unkown return code from the BPF program would fail in tcf_bpf() with TC_ACT_UNSPEC. Should we one day want to make use of TC_ACT_STOLEN or TC_ACT_QUEUED, which both have the same semantics, we have the option to either use that as a default action (filter_res of 0xffffffff) or non-default BPF return code. All that will allow us to transparently use tcf_bpf() for both BPF flavours. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Cc: Jiri Pirko <jiri@resnulli.us> Cc: Alexei Starovoitov <ast@plumgrid.com> Cc: Jamal Hadi Salim <jhs@mojatatu.com> Acked-by: Jiri Pirko <jiri@resnulli.us> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-03-17 19:25:57 +00:00
break;
case TC_ACT_UNSPEC:
action = prog->tcf_action;
act_bpf: allow non-default TC_ACT opcodes as BPF exec outcome Revisiting commit d23b8ad8ab23 ("tc: add BPF based action") with regards to eBPF support, I was thinking that it might be better to improve return semantics from a BPF program invoked through BPF_PROG_RUN(). Currently, in case filter_res is 0, we overwrite the default action opcode with TC_ACT_SHOT. A default action opcode configured through tc's m_bpf can be: TC_ACT_RECLASSIFY, TC_ACT_PIPE, TC_ACT_SHOT, TC_ACT_UNSPEC, TC_ACT_OK. In cls_bpf, we have the possibility to overwrite the default class associated with the classifier in case filter_res is _not_ 0xffffffff (-1). That allows us to fold multiple [e]BPF programs into a single one, where they would otherwise need to be defined as a separate classifier with its own classid, needlessly redoing parsing work, etc. Similarly, we could do better in act_bpf: Since above TC_ACT* opcodes are exported to UAPI anyway, we reuse them for return-code-to-tc-opcode mapping, where we would allow above possibilities. Thus, like in cls_bpf, a filter_res of 0xffffffff (-1) means that the configured _default_ action is used. Any unkown return code from the BPF program would fail in tcf_bpf() with TC_ACT_UNSPEC. Should we one day want to make use of TC_ACT_STOLEN or TC_ACT_QUEUED, which both have the same semantics, we have the option to either use that as a default action (filter_res of 0xffffffff) or non-default BPF return code. All that will allow us to transparently use tcf_bpf() for both BPF flavours. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Cc: Jiri Pirko <jiri@resnulli.us> Cc: Alexei Starovoitov <ast@plumgrid.com> Cc: Jamal Hadi Salim <jhs@mojatatu.com> Acked-by: Jiri Pirko <jiri@resnulli.us> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-03-17 19:25:57 +00:00
break;
default:
action = TC_ACT_UNSPEC;
break;
}
return action;
}
static bool tcf_bpf_is_ebpf(const struct tcf_bpf *prog)
{
return !prog->bpf_ops;
}
static int tcf_bpf_dump_bpf_info(const struct tcf_bpf *prog,
struct sk_buff *skb)
{
struct nlattr *nla;
if (nla_put_u16(skb, TCA_ACT_BPF_OPS_LEN, prog->bpf_num_ops))
return -EMSGSIZE;
nla = nla_reserve(skb, TCA_ACT_BPF_OPS, prog->bpf_num_ops *
sizeof(struct sock_filter));
if (nla == NULL)
return -EMSGSIZE;
memcpy(nla_data(nla), prog->bpf_ops, nla_len(nla));
return 0;
}
static int tcf_bpf_dump_ebpf_info(const struct tcf_bpf *prog,
struct sk_buff *skb)
{
if (nla_put_u32(skb, TCA_ACT_BPF_FD, prog->bpf_fd))
return -EMSGSIZE;
if (prog->bpf_name &&
nla_put_string(skb, TCA_ACT_BPF_NAME, prog->bpf_name))
return -EMSGSIZE;
return 0;
}
static int tcf_bpf_dump(struct sk_buff *skb, struct tc_action *act,
int bind, int ref)
{
unsigned char *tp = skb_tail_pointer(skb);
struct tcf_bpf *prog = act->priv;
struct tc_act_bpf opt = {
.index = prog->tcf_index,
.refcnt = prog->tcf_refcnt - ref,
.bindcnt = prog->tcf_bindcnt - bind,
.action = prog->tcf_action,
};
struct tcf_t tm;
int ret;
if (nla_put(skb, TCA_ACT_BPF_PARMS, sizeof(opt), &opt))
goto nla_put_failure;
if (tcf_bpf_is_ebpf(prog))
ret = tcf_bpf_dump_ebpf_info(prog, skb);
else
ret = tcf_bpf_dump_bpf_info(prog, skb);
if (ret)
goto nla_put_failure;
tm.install = jiffies_to_clock_t(jiffies - prog->tcf_tm.install);
tm.lastuse = jiffies_to_clock_t(jiffies - prog->tcf_tm.lastuse);
tm.expires = jiffies_to_clock_t(prog->tcf_tm.expires);
if (nla_put_64bit(skb, TCA_ACT_BPF_TM, sizeof(tm), &tm,
TCA_ACT_BPF_PAD))
goto nla_put_failure;
return skb->len;
nla_put_failure:
nlmsg_trim(skb, tp);
return -1;
}
static const struct nla_policy act_bpf_policy[TCA_ACT_BPF_MAX + 1] = {
[TCA_ACT_BPF_PARMS] = { .len = sizeof(struct tc_act_bpf) },
[TCA_ACT_BPF_FD] = { .type = NLA_U32 },
[TCA_ACT_BPF_NAME] = { .type = NLA_NUL_STRING, .len = ACT_BPF_NAME_LEN },
[TCA_ACT_BPF_OPS_LEN] = { .type = NLA_U16 },
[TCA_ACT_BPF_OPS] = { .type = NLA_BINARY,
.len = sizeof(struct sock_filter) * BPF_MAXINSNS },
};
static int tcf_bpf_init_from_ops(struct nlattr **tb, struct tcf_bpf_cfg *cfg)
{
struct sock_filter *bpf_ops;
struct sock_fprog_kern fprog_tmp;
struct bpf_prog *fp;
u16 bpf_size, bpf_num_ops;
int ret;
bpf_num_ops = nla_get_u16(tb[TCA_ACT_BPF_OPS_LEN]);
if (bpf_num_ops > BPF_MAXINSNS || bpf_num_ops == 0)
return -EINVAL;
bpf_size = bpf_num_ops * sizeof(*bpf_ops);
if (bpf_size != nla_len(tb[TCA_ACT_BPF_OPS]))
return -EINVAL;
bpf_ops = kzalloc(bpf_size, GFP_KERNEL);
if (bpf_ops == NULL)
return -ENOMEM;
memcpy(bpf_ops, nla_data(tb[TCA_ACT_BPF_OPS]), bpf_size);
fprog_tmp.len = bpf_num_ops;
fprog_tmp.filter = bpf_ops;
ret = bpf_prog_create(&fp, &fprog_tmp);
if (ret < 0) {
kfree(bpf_ops);
return ret;
}
cfg->bpf_ops = bpf_ops;
cfg->bpf_num_ops = bpf_num_ops;
cfg->filter = fp;
act_bpf: fix memory leaks when replacing bpf programs We currently trigger multiple memory leaks when replacing bpf actions, besides others: comm "tc", pid 1909, jiffies 4294851310 (age 1602.796s) hex dump (first 32 bytes): 01 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 ................ 18 b0 98 6d 00 88 ff ff 00 00 00 00 00 00 00 00 ...m............ backtrace: [<ffffffff817e623e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff8120a22d>] __vmalloc_node_range+0x1bd/0x2c0 [<ffffffff8120a37a>] __vmalloc+0x4a/0x50 [<ffffffff811a8d0a>] bpf_prog_alloc+0x3a/0xa0 [<ffffffff816c0684>] bpf_prog_create+0x44/0xa0 [<ffffffffa09ba4eb>] tcf_bpf_init+0x28b/0x3c0 [act_bpf] [<ffffffff816d7001>] tcf_action_init_1+0x191/0x1b0 [<ffffffff816d70a2>] tcf_action_init+0x82/0xf0 [<ffffffff816d4d12>] tcf_exts_validate+0xb2/0xc0 [<ffffffffa09b5838>] cls_bpf_modify_existing+0x98/0x340 [cls_bpf] [<ffffffffa09b5cd6>] cls_bpf_change+0x1a6/0x274 [cls_bpf] [<ffffffff816d56e5>] tc_ctl_tfilter+0x335/0x910 [<ffffffff816b9145>] rtnetlink_rcv_msg+0x95/0x240 [<ffffffff816df34f>] netlink_rcv_skb+0xaf/0xc0 [<ffffffff816b909e>] rtnetlink_rcv+0x2e/0x40 [<ffffffff816deaaf>] netlink_unicast+0xef/0x1b0 Issue is that the old content from tcf_bpf is allocated and needs to be released when we replace it. We seem to do that since the beginning of act_bpf on the filter and insns, later on the name as well. Example test case, after patch: # FOO="1,6 0 0 4294967295," # BAR="1,6 0 0 4294967294," # tc actions add action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$BAR" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967294' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions del action bpf index 2 [...] # echo "scan" > /sys/kernel/debug/kmemleak # cat /sys/kernel/debug/kmemleak | grep "comm \"tc\"" | wc -l 0 Fixes: d23b8ad8ab23 ("tc: add BPF based action") Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-07-29 16:40:56 +00:00
cfg->is_ebpf = false;
return 0;
}
static int tcf_bpf_init_from_efd(struct nlattr **tb, struct tcf_bpf_cfg *cfg)
{
struct bpf_prog *fp;
char *name = NULL;
u32 bpf_fd;
bpf_fd = nla_get_u32(tb[TCA_ACT_BPF_FD]);
fp = bpf_prog_get(bpf_fd);
if (IS_ERR(fp))
return PTR_ERR(fp);
if (fp->type != BPF_PROG_TYPE_SCHED_ACT) {
bpf_prog_put(fp);
return -EINVAL;
}
if (tb[TCA_ACT_BPF_NAME]) {
name = kmemdup(nla_data(tb[TCA_ACT_BPF_NAME]),
nla_len(tb[TCA_ACT_BPF_NAME]),
GFP_KERNEL);
if (!name) {
bpf_prog_put(fp);
return -ENOMEM;
}
}
cfg->bpf_fd = bpf_fd;
cfg->bpf_name = name;
cfg->filter = fp;
act_bpf: fix memory leaks when replacing bpf programs We currently trigger multiple memory leaks when replacing bpf actions, besides others: comm "tc", pid 1909, jiffies 4294851310 (age 1602.796s) hex dump (first 32 bytes): 01 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 ................ 18 b0 98 6d 00 88 ff ff 00 00 00 00 00 00 00 00 ...m............ backtrace: [<ffffffff817e623e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff8120a22d>] __vmalloc_node_range+0x1bd/0x2c0 [<ffffffff8120a37a>] __vmalloc+0x4a/0x50 [<ffffffff811a8d0a>] bpf_prog_alloc+0x3a/0xa0 [<ffffffff816c0684>] bpf_prog_create+0x44/0xa0 [<ffffffffa09ba4eb>] tcf_bpf_init+0x28b/0x3c0 [act_bpf] [<ffffffff816d7001>] tcf_action_init_1+0x191/0x1b0 [<ffffffff816d70a2>] tcf_action_init+0x82/0xf0 [<ffffffff816d4d12>] tcf_exts_validate+0xb2/0xc0 [<ffffffffa09b5838>] cls_bpf_modify_existing+0x98/0x340 [cls_bpf] [<ffffffffa09b5cd6>] cls_bpf_change+0x1a6/0x274 [cls_bpf] [<ffffffff816d56e5>] tc_ctl_tfilter+0x335/0x910 [<ffffffff816b9145>] rtnetlink_rcv_msg+0x95/0x240 [<ffffffff816df34f>] netlink_rcv_skb+0xaf/0xc0 [<ffffffff816b909e>] rtnetlink_rcv+0x2e/0x40 [<ffffffff816deaaf>] netlink_unicast+0xef/0x1b0 Issue is that the old content from tcf_bpf is allocated and needs to be released when we replace it. We seem to do that since the beginning of act_bpf on the filter and insns, later on the name as well. Example test case, after patch: # FOO="1,6 0 0 4294967295," # BAR="1,6 0 0 4294967294," # tc actions add action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$BAR" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967294' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions del action bpf index 2 [...] # echo "scan" > /sys/kernel/debug/kmemleak # cat /sys/kernel/debug/kmemleak | grep "comm \"tc\"" | wc -l 0 Fixes: d23b8ad8ab23 ("tc: add BPF based action") Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-07-29 16:40:56 +00:00
cfg->is_ebpf = true;
return 0;
}
act_bpf: fix memory leaks when replacing bpf programs We currently trigger multiple memory leaks when replacing bpf actions, besides others: comm "tc", pid 1909, jiffies 4294851310 (age 1602.796s) hex dump (first 32 bytes): 01 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 ................ 18 b0 98 6d 00 88 ff ff 00 00 00 00 00 00 00 00 ...m............ backtrace: [<ffffffff817e623e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff8120a22d>] __vmalloc_node_range+0x1bd/0x2c0 [<ffffffff8120a37a>] __vmalloc+0x4a/0x50 [<ffffffff811a8d0a>] bpf_prog_alloc+0x3a/0xa0 [<ffffffff816c0684>] bpf_prog_create+0x44/0xa0 [<ffffffffa09ba4eb>] tcf_bpf_init+0x28b/0x3c0 [act_bpf] [<ffffffff816d7001>] tcf_action_init_1+0x191/0x1b0 [<ffffffff816d70a2>] tcf_action_init+0x82/0xf0 [<ffffffff816d4d12>] tcf_exts_validate+0xb2/0xc0 [<ffffffffa09b5838>] cls_bpf_modify_existing+0x98/0x340 [cls_bpf] [<ffffffffa09b5cd6>] cls_bpf_change+0x1a6/0x274 [cls_bpf] [<ffffffff816d56e5>] tc_ctl_tfilter+0x335/0x910 [<ffffffff816b9145>] rtnetlink_rcv_msg+0x95/0x240 [<ffffffff816df34f>] netlink_rcv_skb+0xaf/0xc0 [<ffffffff816b909e>] rtnetlink_rcv+0x2e/0x40 [<ffffffff816deaaf>] netlink_unicast+0xef/0x1b0 Issue is that the old content from tcf_bpf is allocated and needs to be released when we replace it. We seem to do that since the beginning of act_bpf on the filter and insns, later on the name as well. Example test case, after patch: # FOO="1,6 0 0 4294967295," # BAR="1,6 0 0 4294967294," # tc actions add action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$BAR" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967294' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions del action bpf index 2 [...] # echo "scan" > /sys/kernel/debug/kmemleak # cat /sys/kernel/debug/kmemleak | grep "comm \"tc\"" | wc -l 0 Fixes: d23b8ad8ab23 ("tc: add BPF based action") Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-07-29 16:40:56 +00:00
static void tcf_bpf_cfg_cleanup(const struct tcf_bpf_cfg *cfg)
{
if (cfg->is_ebpf)
bpf_prog_put(cfg->filter);
else
bpf_prog_destroy(cfg->filter);
kfree(cfg->bpf_ops);
kfree(cfg->bpf_name);
}
static void tcf_bpf_prog_fill_cfg(const struct tcf_bpf *prog,
struct tcf_bpf_cfg *cfg)
{
cfg->is_ebpf = tcf_bpf_is_ebpf(prog);
/* updates to prog->filter are prevented, since it's called either
* with rtnl lock or during final cleanup in rcu callback
*/
cfg->filter = rcu_dereference_protected(prog->filter, 1);
act_bpf: fix memory leaks when replacing bpf programs We currently trigger multiple memory leaks when replacing bpf actions, besides others: comm "tc", pid 1909, jiffies 4294851310 (age 1602.796s) hex dump (first 32 bytes): 01 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 ................ 18 b0 98 6d 00 88 ff ff 00 00 00 00 00 00 00 00 ...m............ backtrace: [<ffffffff817e623e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff8120a22d>] __vmalloc_node_range+0x1bd/0x2c0 [<ffffffff8120a37a>] __vmalloc+0x4a/0x50 [<ffffffff811a8d0a>] bpf_prog_alloc+0x3a/0xa0 [<ffffffff816c0684>] bpf_prog_create+0x44/0xa0 [<ffffffffa09ba4eb>] tcf_bpf_init+0x28b/0x3c0 [act_bpf] [<ffffffff816d7001>] tcf_action_init_1+0x191/0x1b0 [<ffffffff816d70a2>] tcf_action_init+0x82/0xf0 [<ffffffff816d4d12>] tcf_exts_validate+0xb2/0xc0 [<ffffffffa09b5838>] cls_bpf_modify_existing+0x98/0x340 [cls_bpf] [<ffffffffa09b5cd6>] cls_bpf_change+0x1a6/0x274 [cls_bpf] [<ffffffff816d56e5>] tc_ctl_tfilter+0x335/0x910 [<ffffffff816b9145>] rtnetlink_rcv_msg+0x95/0x240 [<ffffffff816df34f>] netlink_rcv_skb+0xaf/0xc0 [<ffffffff816b909e>] rtnetlink_rcv+0x2e/0x40 [<ffffffff816deaaf>] netlink_unicast+0xef/0x1b0 Issue is that the old content from tcf_bpf is allocated and needs to be released when we replace it. We seem to do that since the beginning of act_bpf on the filter and insns, later on the name as well. Example test case, after patch: # FOO="1,6 0 0 4294967295," # BAR="1,6 0 0 4294967294," # tc actions add action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$BAR" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967294' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions del action bpf index 2 [...] # echo "scan" > /sys/kernel/debug/kmemleak # cat /sys/kernel/debug/kmemleak | grep "comm \"tc\"" | wc -l 0 Fixes: d23b8ad8ab23 ("tc: add BPF based action") Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-07-29 16:40:56 +00:00
cfg->bpf_ops = prog->bpf_ops;
cfg->bpf_name = prog->bpf_name;
}
static int tcf_bpf_init(struct net *net, struct nlattr *nla,
struct nlattr *est, struct tc_action *act,
int replace, int bind)
{
struct tc_action_net *tn = net_generic(net, bpf_net_id);
struct nlattr *tb[TCA_ACT_BPF_MAX + 1];
act_bpf: fix memory leaks when replacing bpf programs We currently trigger multiple memory leaks when replacing bpf actions, besides others: comm "tc", pid 1909, jiffies 4294851310 (age 1602.796s) hex dump (first 32 bytes): 01 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 ................ 18 b0 98 6d 00 88 ff ff 00 00 00 00 00 00 00 00 ...m............ backtrace: [<ffffffff817e623e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff8120a22d>] __vmalloc_node_range+0x1bd/0x2c0 [<ffffffff8120a37a>] __vmalloc+0x4a/0x50 [<ffffffff811a8d0a>] bpf_prog_alloc+0x3a/0xa0 [<ffffffff816c0684>] bpf_prog_create+0x44/0xa0 [<ffffffffa09ba4eb>] tcf_bpf_init+0x28b/0x3c0 [act_bpf] [<ffffffff816d7001>] tcf_action_init_1+0x191/0x1b0 [<ffffffff816d70a2>] tcf_action_init+0x82/0xf0 [<ffffffff816d4d12>] tcf_exts_validate+0xb2/0xc0 [<ffffffffa09b5838>] cls_bpf_modify_existing+0x98/0x340 [cls_bpf] [<ffffffffa09b5cd6>] cls_bpf_change+0x1a6/0x274 [cls_bpf] [<ffffffff816d56e5>] tc_ctl_tfilter+0x335/0x910 [<ffffffff816b9145>] rtnetlink_rcv_msg+0x95/0x240 [<ffffffff816df34f>] netlink_rcv_skb+0xaf/0xc0 [<ffffffff816b909e>] rtnetlink_rcv+0x2e/0x40 [<ffffffff816deaaf>] netlink_unicast+0xef/0x1b0 Issue is that the old content from tcf_bpf is allocated and needs to be released when we replace it. We seem to do that since the beginning of act_bpf on the filter and insns, later on the name as well. Example test case, after patch: # FOO="1,6 0 0 4294967295," # BAR="1,6 0 0 4294967294," # tc actions add action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$BAR" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967294' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions del action bpf index 2 [...] # echo "scan" > /sys/kernel/debug/kmemleak # cat /sys/kernel/debug/kmemleak | grep "comm \"tc\"" | wc -l 0 Fixes: d23b8ad8ab23 ("tc: add BPF based action") Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-07-29 16:40:56 +00:00
struct tcf_bpf_cfg cfg, old;
struct tc_act_bpf *parm;
struct tcf_bpf *prog;
bool is_bpf, is_ebpf;
act_bpf: properly support late binding of bpf action to a classifier Since the introduction of the BPF action in d23b8ad8ab23 ("tc: add BPF based action"), late binding was not working as expected. I.e. setting the action part for a classifier only via 'bpf index <num>', where <num> is the index of an existing action, is being rejected by the kernel due to other missing parameters. It doesn't make sense to require these parameters such as BPF opcodes etc, as they are not going to be used anyway: in this case, they're just allocated/parsed and then freed again w/o doing anything meaningful. Instead, parse and verify the remaining parameters *after* the test on tcf_hash_check(), when we really know that we're dealing with creation of a new action or replacement of an existing one and where late binding is thus irrelevant. After patch, test case is now working: FOO="1,6 0 0 4294967295," tc actions add action bpf bytecode "$FOO" tc filter add dev foo parent 1: bpf bytecode "$FOO" flowid 1:1 action bpf index 1 tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 tc filter show dev foo filter protocol all pref 49152 bpf filter protocol all pref 49152 bpf handle 0x1 flowid 1:1 bytecode '1,6 0 0 4294967295' action order 1: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 Late binding of a BPF action can be useful for preloading maps (e.g. before they hit traffic) in case of eBPF programs, or to share a single eBPF action with multiple classifiers. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Alexei Starovoitov <ast@plumgrid.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-08-03 14:21:57 +00:00
int ret, res = 0;
if (!nla)
return -EINVAL;
ret = nla_parse_nested(tb, TCA_ACT_BPF_MAX, nla, act_bpf_policy);
if (ret < 0)
return ret;
act_bpf: properly support late binding of bpf action to a classifier Since the introduction of the BPF action in d23b8ad8ab23 ("tc: add BPF based action"), late binding was not working as expected. I.e. setting the action part for a classifier only via 'bpf index <num>', where <num> is the index of an existing action, is being rejected by the kernel due to other missing parameters. It doesn't make sense to require these parameters such as BPF opcodes etc, as they are not going to be used anyway: in this case, they're just allocated/parsed and then freed again w/o doing anything meaningful. Instead, parse and verify the remaining parameters *after* the test on tcf_hash_check(), when we really know that we're dealing with creation of a new action or replacement of an existing one and where late binding is thus irrelevant. After patch, test case is now working: FOO="1,6 0 0 4294967295," tc actions add action bpf bytecode "$FOO" tc filter add dev foo parent 1: bpf bytecode "$FOO" flowid 1:1 action bpf index 1 tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 tc filter show dev foo filter protocol all pref 49152 bpf filter protocol all pref 49152 bpf handle 0x1 flowid 1:1 bytecode '1,6 0 0 4294967295' action order 1: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 Late binding of a BPF action can be useful for preloading maps (e.g. before they hit traffic) in case of eBPF programs, or to share a single eBPF action with multiple classifiers. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Alexei Starovoitov <ast@plumgrid.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-08-03 14:21:57 +00:00
if (!tb[TCA_ACT_BPF_PARMS])
return -EINVAL;
parm = nla_data(tb[TCA_ACT_BPF_PARMS]);
if (!tcf_hash_check(tn, parm->index, act, bind)) {
ret = tcf_hash_create(tn, parm->index, est, act,
sizeof(*prog), bind, true);
if (ret < 0)
act_bpf: properly support late binding of bpf action to a classifier Since the introduction of the BPF action in d23b8ad8ab23 ("tc: add BPF based action"), late binding was not working as expected. I.e. setting the action part for a classifier only via 'bpf index <num>', where <num> is the index of an existing action, is being rejected by the kernel due to other missing parameters. It doesn't make sense to require these parameters such as BPF opcodes etc, as they are not going to be used anyway: in this case, they're just allocated/parsed and then freed again w/o doing anything meaningful. Instead, parse and verify the remaining parameters *after* the test on tcf_hash_check(), when we really know that we're dealing with creation of a new action or replacement of an existing one and where late binding is thus irrelevant. After patch, test case is now working: FOO="1,6 0 0 4294967295," tc actions add action bpf bytecode "$FOO" tc filter add dev foo parent 1: bpf bytecode "$FOO" flowid 1:1 action bpf index 1 tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 tc filter show dev foo filter protocol all pref 49152 bpf filter protocol all pref 49152 bpf handle 0x1 flowid 1:1 bytecode '1,6 0 0 4294967295' action order 1: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 Late binding of a BPF action can be useful for preloading maps (e.g. before they hit traffic) in case of eBPF programs, or to share a single eBPF action with multiple classifiers. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Alexei Starovoitov <ast@plumgrid.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-08-03 14:21:57 +00:00
return ret;
act_bpf: properly support late binding of bpf action to a classifier Since the introduction of the BPF action in d23b8ad8ab23 ("tc: add BPF based action"), late binding was not working as expected. I.e. setting the action part for a classifier only via 'bpf index <num>', where <num> is the index of an existing action, is being rejected by the kernel due to other missing parameters. It doesn't make sense to require these parameters such as BPF opcodes etc, as they are not going to be used anyway: in this case, they're just allocated/parsed and then freed again w/o doing anything meaningful. Instead, parse and verify the remaining parameters *after* the test on tcf_hash_check(), when we really know that we're dealing with creation of a new action or replacement of an existing one and where late binding is thus irrelevant. After patch, test case is now working: FOO="1,6 0 0 4294967295," tc actions add action bpf bytecode "$FOO" tc filter add dev foo parent 1: bpf bytecode "$FOO" flowid 1:1 action bpf index 1 tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 tc filter show dev foo filter protocol all pref 49152 bpf filter protocol all pref 49152 bpf handle 0x1 flowid 1:1 bytecode '1,6 0 0 4294967295' action order 1: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 Late binding of a BPF action can be useful for preloading maps (e.g. before they hit traffic) in case of eBPF programs, or to share a single eBPF action with multiple classifiers. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Alexei Starovoitov <ast@plumgrid.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-08-03 14:21:57 +00:00
res = ACT_P_CREATED;
} else {
/* Don't override defaults. */
if (bind)
act_bpf: properly support late binding of bpf action to a classifier Since the introduction of the BPF action in d23b8ad8ab23 ("tc: add BPF based action"), late binding was not working as expected. I.e. setting the action part for a classifier only via 'bpf index <num>', where <num> is the index of an existing action, is being rejected by the kernel due to other missing parameters. It doesn't make sense to require these parameters such as BPF opcodes etc, as they are not going to be used anyway: in this case, they're just allocated/parsed and then freed again w/o doing anything meaningful. Instead, parse and verify the remaining parameters *after* the test on tcf_hash_check(), when we really know that we're dealing with creation of a new action or replacement of an existing one and where late binding is thus irrelevant. After patch, test case is now working: FOO="1,6 0 0 4294967295," tc actions add action bpf bytecode "$FOO" tc filter add dev foo parent 1: bpf bytecode "$FOO" flowid 1:1 action bpf index 1 tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 tc filter show dev foo filter protocol all pref 49152 bpf filter protocol all pref 49152 bpf handle 0x1 flowid 1:1 bytecode '1,6 0 0 4294967295' action order 1: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 Late binding of a BPF action can be useful for preloading maps (e.g. before they hit traffic) in case of eBPF programs, or to share a single eBPF action with multiple classifiers. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Alexei Starovoitov <ast@plumgrid.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-08-03 14:21:57 +00:00
return 0;
tcf_hash_release(act, bind);
act_bpf: properly support late binding of bpf action to a classifier Since the introduction of the BPF action in d23b8ad8ab23 ("tc: add BPF based action"), late binding was not working as expected. I.e. setting the action part for a classifier only via 'bpf index <num>', where <num> is the index of an existing action, is being rejected by the kernel due to other missing parameters. It doesn't make sense to require these parameters such as BPF opcodes etc, as they are not going to be used anyway: in this case, they're just allocated/parsed and then freed again w/o doing anything meaningful. Instead, parse and verify the remaining parameters *after* the test on tcf_hash_check(), when we really know that we're dealing with creation of a new action or replacement of an existing one and where late binding is thus irrelevant. After patch, test case is now working: FOO="1,6 0 0 4294967295," tc actions add action bpf bytecode "$FOO" tc filter add dev foo parent 1: bpf bytecode "$FOO" flowid 1:1 action bpf index 1 tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 tc filter show dev foo filter protocol all pref 49152 bpf filter protocol all pref 49152 bpf handle 0x1 flowid 1:1 bytecode '1,6 0 0 4294967295' action order 1: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 Late binding of a BPF action can be useful for preloading maps (e.g. before they hit traffic) in case of eBPF programs, or to share a single eBPF action with multiple classifiers. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Alexei Starovoitov <ast@plumgrid.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-08-03 14:21:57 +00:00
if (!replace)
return -EEXIST;
}
act_bpf: properly support late binding of bpf action to a classifier Since the introduction of the BPF action in d23b8ad8ab23 ("tc: add BPF based action"), late binding was not working as expected. I.e. setting the action part for a classifier only via 'bpf index <num>', where <num> is the index of an existing action, is being rejected by the kernel due to other missing parameters. It doesn't make sense to require these parameters such as BPF opcodes etc, as they are not going to be used anyway: in this case, they're just allocated/parsed and then freed again w/o doing anything meaningful. Instead, parse and verify the remaining parameters *after* the test on tcf_hash_check(), when we really know that we're dealing with creation of a new action or replacement of an existing one and where late binding is thus irrelevant. After patch, test case is now working: FOO="1,6 0 0 4294967295," tc actions add action bpf bytecode "$FOO" tc filter add dev foo parent 1: bpf bytecode "$FOO" flowid 1:1 action bpf index 1 tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 tc filter show dev foo filter protocol all pref 49152 bpf filter protocol all pref 49152 bpf handle 0x1 flowid 1:1 bytecode '1,6 0 0 4294967295' action order 1: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 Late binding of a BPF action can be useful for preloading maps (e.g. before they hit traffic) in case of eBPF programs, or to share a single eBPF action with multiple classifiers. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Alexei Starovoitov <ast@plumgrid.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-08-03 14:21:57 +00:00
is_bpf = tb[TCA_ACT_BPF_OPS_LEN] && tb[TCA_ACT_BPF_OPS];
is_ebpf = tb[TCA_ACT_BPF_FD];
if ((!is_bpf && !is_ebpf) || (is_bpf && is_ebpf)) {
ret = -EINVAL;
goto out;
}
memset(&cfg, 0, sizeof(cfg));
ret = is_bpf ? tcf_bpf_init_from_ops(tb, &cfg) :
tcf_bpf_init_from_efd(tb, &cfg);
if (ret < 0)
goto out;
prog = to_bpf(act);
ASSERT_RTNL();
if (res != ACT_P_CREATED)
act_bpf: fix memory leaks when replacing bpf programs We currently trigger multiple memory leaks when replacing bpf actions, besides others: comm "tc", pid 1909, jiffies 4294851310 (age 1602.796s) hex dump (first 32 bytes): 01 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 ................ 18 b0 98 6d 00 88 ff ff 00 00 00 00 00 00 00 00 ...m............ backtrace: [<ffffffff817e623e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff8120a22d>] __vmalloc_node_range+0x1bd/0x2c0 [<ffffffff8120a37a>] __vmalloc+0x4a/0x50 [<ffffffff811a8d0a>] bpf_prog_alloc+0x3a/0xa0 [<ffffffff816c0684>] bpf_prog_create+0x44/0xa0 [<ffffffffa09ba4eb>] tcf_bpf_init+0x28b/0x3c0 [act_bpf] [<ffffffff816d7001>] tcf_action_init_1+0x191/0x1b0 [<ffffffff816d70a2>] tcf_action_init+0x82/0xf0 [<ffffffff816d4d12>] tcf_exts_validate+0xb2/0xc0 [<ffffffffa09b5838>] cls_bpf_modify_existing+0x98/0x340 [cls_bpf] [<ffffffffa09b5cd6>] cls_bpf_change+0x1a6/0x274 [cls_bpf] [<ffffffff816d56e5>] tc_ctl_tfilter+0x335/0x910 [<ffffffff816b9145>] rtnetlink_rcv_msg+0x95/0x240 [<ffffffff816df34f>] netlink_rcv_skb+0xaf/0xc0 [<ffffffff816b909e>] rtnetlink_rcv+0x2e/0x40 [<ffffffff816deaaf>] netlink_unicast+0xef/0x1b0 Issue is that the old content from tcf_bpf is allocated and needs to be released when we replace it. We seem to do that since the beginning of act_bpf on the filter and insns, later on the name as well. Example test case, after patch: # FOO="1,6 0 0 4294967295," # BAR="1,6 0 0 4294967294," # tc actions add action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$BAR" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967294' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions del action bpf index 2 [...] # echo "scan" > /sys/kernel/debug/kmemleak # cat /sys/kernel/debug/kmemleak | grep "comm \"tc\"" | wc -l 0 Fixes: d23b8ad8ab23 ("tc: add BPF based action") Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-07-29 16:40:56 +00:00
tcf_bpf_prog_fill_cfg(prog, &old);
prog->bpf_ops = cfg.bpf_ops;
prog->bpf_name = cfg.bpf_name;
if (cfg.bpf_num_ops)
prog->bpf_num_ops = cfg.bpf_num_ops;
if (cfg.bpf_fd)
prog->bpf_fd = cfg.bpf_fd;
prog->tcf_action = parm->action;
rcu_assign_pointer(prog->filter, cfg.filter);
if (res == ACT_P_CREATED) {
tcf_hash_insert(tn, act);
} else {
/* make sure the program being replaced is no longer executing */
synchronize_rcu();
act_bpf: fix memory leaks when replacing bpf programs We currently trigger multiple memory leaks when replacing bpf actions, besides others: comm "tc", pid 1909, jiffies 4294851310 (age 1602.796s) hex dump (first 32 bytes): 01 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 ................ 18 b0 98 6d 00 88 ff ff 00 00 00 00 00 00 00 00 ...m............ backtrace: [<ffffffff817e623e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff8120a22d>] __vmalloc_node_range+0x1bd/0x2c0 [<ffffffff8120a37a>] __vmalloc+0x4a/0x50 [<ffffffff811a8d0a>] bpf_prog_alloc+0x3a/0xa0 [<ffffffff816c0684>] bpf_prog_create+0x44/0xa0 [<ffffffffa09ba4eb>] tcf_bpf_init+0x28b/0x3c0 [act_bpf] [<ffffffff816d7001>] tcf_action_init_1+0x191/0x1b0 [<ffffffff816d70a2>] tcf_action_init+0x82/0xf0 [<ffffffff816d4d12>] tcf_exts_validate+0xb2/0xc0 [<ffffffffa09b5838>] cls_bpf_modify_existing+0x98/0x340 [cls_bpf] [<ffffffffa09b5cd6>] cls_bpf_change+0x1a6/0x274 [cls_bpf] [<ffffffff816d56e5>] tc_ctl_tfilter+0x335/0x910 [<ffffffff816b9145>] rtnetlink_rcv_msg+0x95/0x240 [<ffffffff816df34f>] netlink_rcv_skb+0xaf/0xc0 [<ffffffff816b909e>] rtnetlink_rcv+0x2e/0x40 [<ffffffff816deaaf>] netlink_unicast+0xef/0x1b0 Issue is that the old content from tcf_bpf is allocated and needs to be released when we replace it. We seem to do that since the beginning of act_bpf on the filter and insns, later on the name as well. Example test case, after patch: # FOO="1,6 0 0 4294967295," # BAR="1,6 0 0 4294967294," # tc actions add action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$BAR" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967294' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions del action bpf index 2 [...] # echo "scan" > /sys/kernel/debug/kmemleak # cat /sys/kernel/debug/kmemleak | grep "comm \"tc\"" | wc -l 0 Fixes: d23b8ad8ab23 ("tc: add BPF based action") Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-07-29 16:40:56 +00:00
tcf_bpf_cfg_cleanup(&old);
}
act_bpf: properly support late binding of bpf action to a classifier Since the introduction of the BPF action in d23b8ad8ab23 ("tc: add BPF based action"), late binding was not working as expected. I.e. setting the action part for a classifier only via 'bpf index <num>', where <num> is the index of an existing action, is being rejected by the kernel due to other missing parameters. It doesn't make sense to require these parameters such as BPF opcodes etc, as they are not going to be used anyway: in this case, they're just allocated/parsed and then freed again w/o doing anything meaningful. Instead, parse and verify the remaining parameters *after* the test on tcf_hash_check(), when we really know that we're dealing with creation of a new action or replacement of an existing one and where late binding is thus irrelevant. After patch, test case is now working: FOO="1,6 0 0 4294967295," tc actions add action bpf bytecode "$FOO" tc filter add dev foo parent 1: bpf bytecode "$FOO" flowid 1:1 action bpf index 1 tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 tc filter show dev foo filter protocol all pref 49152 bpf filter protocol all pref 49152 bpf handle 0x1 flowid 1:1 bytecode '1,6 0 0 4294967295' action order 1: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 1 ref 2 bind 1 Late binding of a BPF action can be useful for preloading maps (e.g. before they hit traffic) in case of eBPF programs, or to share a single eBPF action with multiple classifiers. Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Alexei Starovoitov <ast@plumgrid.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-08-03 14:21:57 +00:00
return res;
out:
if (res == ACT_P_CREATED)
tcf_hash_cleanup(act, est);
return ret;
}
static void tcf_bpf_cleanup(struct tc_action *act, int bind)
{
act_bpf: fix memory leaks when replacing bpf programs We currently trigger multiple memory leaks when replacing bpf actions, besides others: comm "tc", pid 1909, jiffies 4294851310 (age 1602.796s) hex dump (first 32 bytes): 01 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 ................ 18 b0 98 6d 00 88 ff ff 00 00 00 00 00 00 00 00 ...m............ backtrace: [<ffffffff817e623e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff8120a22d>] __vmalloc_node_range+0x1bd/0x2c0 [<ffffffff8120a37a>] __vmalloc+0x4a/0x50 [<ffffffff811a8d0a>] bpf_prog_alloc+0x3a/0xa0 [<ffffffff816c0684>] bpf_prog_create+0x44/0xa0 [<ffffffffa09ba4eb>] tcf_bpf_init+0x28b/0x3c0 [act_bpf] [<ffffffff816d7001>] tcf_action_init_1+0x191/0x1b0 [<ffffffff816d70a2>] tcf_action_init+0x82/0xf0 [<ffffffff816d4d12>] tcf_exts_validate+0xb2/0xc0 [<ffffffffa09b5838>] cls_bpf_modify_existing+0x98/0x340 [cls_bpf] [<ffffffffa09b5cd6>] cls_bpf_change+0x1a6/0x274 [cls_bpf] [<ffffffff816d56e5>] tc_ctl_tfilter+0x335/0x910 [<ffffffff816b9145>] rtnetlink_rcv_msg+0x95/0x240 [<ffffffff816df34f>] netlink_rcv_skb+0xaf/0xc0 [<ffffffff816b909e>] rtnetlink_rcv+0x2e/0x40 [<ffffffff816deaaf>] netlink_unicast+0xef/0x1b0 Issue is that the old content from tcf_bpf is allocated and needs to be released when we replace it. We seem to do that since the beginning of act_bpf on the filter and insns, later on the name as well. Example test case, after patch: # FOO="1,6 0 0 4294967295," # BAR="1,6 0 0 4294967294," # tc actions add action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$BAR" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967294' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions del action bpf index 2 [...] # echo "scan" > /sys/kernel/debug/kmemleak # cat /sys/kernel/debug/kmemleak | grep "comm \"tc\"" | wc -l 0 Fixes: d23b8ad8ab23 ("tc: add BPF based action") Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-07-29 16:40:56 +00:00
struct tcf_bpf_cfg tmp;
act_bpf: fix memory leaks when replacing bpf programs We currently trigger multiple memory leaks when replacing bpf actions, besides others: comm "tc", pid 1909, jiffies 4294851310 (age 1602.796s) hex dump (first 32 bytes): 01 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 ................ 18 b0 98 6d 00 88 ff ff 00 00 00 00 00 00 00 00 ...m............ backtrace: [<ffffffff817e623e>] kmemleak_alloc+0x4e/0xb0 [<ffffffff8120a22d>] __vmalloc_node_range+0x1bd/0x2c0 [<ffffffff8120a37a>] __vmalloc+0x4a/0x50 [<ffffffff811a8d0a>] bpf_prog_alloc+0x3a/0xa0 [<ffffffff816c0684>] bpf_prog_create+0x44/0xa0 [<ffffffffa09ba4eb>] tcf_bpf_init+0x28b/0x3c0 [act_bpf] [<ffffffff816d7001>] tcf_action_init_1+0x191/0x1b0 [<ffffffff816d70a2>] tcf_action_init+0x82/0xf0 [<ffffffff816d4d12>] tcf_exts_validate+0xb2/0xc0 [<ffffffffa09b5838>] cls_bpf_modify_existing+0x98/0x340 [cls_bpf] [<ffffffffa09b5cd6>] cls_bpf_change+0x1a6/0x274 [cls_bpf] [<ffffffff816d56e5>] tc_ctl_tfilter+0x335/0x910 [<ffffffff816b9145>] rtnetlink_rcv_msg+0x95/0x240 [<ffffffff816df34f>] netlink_rcv_skb+0xaf/0xc0 [<ffffffff816b909e>] rtnetlink_rcv+0x2e/0x40 [<ffffffff816deaaf>] netlink_unicast+0xef/0x1b0 Issue is that the old content from tcf_bpf is allocated and needs to be released when we replace it. We seem to do that since the beginning of act_bpf on the filter and insns, later on the name as well. Example test case, after patch: # FOO="1,6 0 0 4294967295," # BAR="1,6 0 0 4294967294," # tc actions add action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$BAR" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967294' default-action pipe index 2 ref 1 bind 0 # tc actions replace action bpf bytecode "$FOO" index 2 # tc actions show action bpf action order 0: bpf bytecode '1,6 0 0 4294967295' default-action pipe index 2 ref 1 bind 0 # tc actions del action bpf index 2 [...] # echo "scan" > /sys/kernel/debug/kmemleak # cat /sys/kernel/debug/kmemleak | grep "comm \"tc\"" | wc -l 0 Fixes: d23b8ad8ab23 ("tc: add BPF based action") Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-07-29 16:40:56 +00:00
tcf_bpf_prog_fill_cfg(act->priv, &tmp);
tcf_bpf_cfg_cleanup(&tmp);
}
static int tcf_bpf_walker(struct net *net, struct sk_buff *skb,
struct netlink_callback *cb, int type,
struct tc_action *a)
{
struct tc_action_net *tn = net_generic(net, bpf_net_id);
return tcf_generic_walker(tn, skb, cb, type, a);
}
static int tcf_bpf_search(struct net *net, struct tc_action *a, u32 index)
{
struct tc_action_net *tn = net_generic(net, bpf_net_id);
return tcf_hash_search(tn, a, index);
}
static struct tc_action_ops act_bpf_ops __read_mostly = {
.kind = "bpf",
.type = TCA_ACT_BPF,
.owner = THIS_MODULE,
.act = tcf_bpf,
.dump = tcf_bpf_dump,
.cleanup = tcf_bpf_cleanup,
.init = tcf_bpf_init,
.walk = tcf_bpf_walker,
.lookup = tcf_bpf_search,
};
static __net_init int bpf_init_net(struct net *net)
{
struct tc_action_net *tn = net_generic(net, bpf_net_id);
return tc_action_net_init(tn, &act_bpf_ops, BPF_TAB_MASK);
}
static void __net_exit bpf_exit_net(struct net *net)
{
struct tc_action_net *tn = net_generic(net, bpf_net_id);
tc_action_net_exit(tn);
}
static struct pernet_operations bpf_net_ops = {
.init = bpf_init_net,
.exit = bpf_exit_net,
.id = &bpf_net_id,
.size = sizeof(struct tc_action_net),
};
static int __init bpf_init_module(void)
{
return tcf_register_action(&act_bpf_ops, &bpf_net_ops);
}
static void __exit bpf_cleanup_module(void)
{
tcf_unregister_action(&act_bpf_ops, &bpf_net_ops);
}
module_init(bpf_init_module);
module_exit(bpf_cleanup_module);
MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
MODULE_DESCRIPTION("TC BPF based action");
MODULE_LICENSE("GPL v2");