You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
suricata/src/flow.h

737 lines
24 KiB
C

/* Copyright (C) 2007-2013 Open Information Security Foundation
*
* You can copy, redistribute or modify this Program under the terms of
* the GNU General Public License version 2 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.
*
* You should have received a copy of the GNU General Public License
* version 2 along with this program; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
* 02110-1301, USA.
*/
/**
* \file
*
* \author Victor Julien <victor@inliniac.net>
*/
#ifndef __FLOW_H__
#define __FLOW_H__
/* forward declaration for macset include */
typedef struct FlowStorageId FlowStorageId;
#include "decode.h"
#include "util-exception-policy.h"
#include "util-var.h"
#include "util-optimize.h"
#include "app-layer-protos.h"
/* Part of the flow structure, so we declare it here.
* The actual declaration is in app-layer-parser.c */
typedef struct AppLayerParserState_ AppLayerParserState;
#define FLOW_QUIET true
#define FLOW_VERBOSE false
#define TOSERVER 0
#define TOCLIENT 1
/* per flow flags */
/** At least one packet from the source address was seen */
#define FLOW_TO_SRC_SEEN BIT_U32(0)
/** At least one packet from the destination address was seen */
#define FLOW_TO_DST_SEEN BIT_U32(1)
flow: handle TCP session reuse in flow engine Until now, TCP session reuse was handled in the TCP stream engine. If the state was TCP_CLOSED, a new SYN packet was received and a few other conditions were met, the flow was 'reset' and reused for the 'new' TCP session. There are a number of problems with this approach: - it breaks the normal flow lifecycle wrt timeout, detection, logging - new TCP sessions could come in on different threads due to mismatches in timeouts between suricata and flow balancing hw/nic/drivers - cleanup code was often causing problems - it complicated locking because of the possible thread mismatch This patch implements a different solution, where a new TCP session also gets a new flow. To do this 2 main things needed to be done: 1. the flow engine needed to be aware of when the TCP reuse case was happening 2. the flow engine needs to be able to 'skip' the old flow once it was replaced by a new one To handle (1), a new function TcpSessionPacketSsnReuse() is introduced to check for the TCP reuse conditions. It's called from 'FlowCompare()' for TCP packets / TCP flows that are candidates for reuse. FlowCompare returns FALSE for the 'old' flow in the case of TCP reuse. This in turn will lead to the flow engine not finding a flow for the TCP SYN packet, resulting in the creation of a new flow. To handle (2), FlowCompare flags the 'old' flow. This flag causes future FlowCompare calls to always return FALSE on it. In other words, the flow can't be found anymore. It can only be accessed by: 1. existing packets with a reference to it 2. flow timeout handling as this logic gets the flows from walking the hash directly 3. flow timeout pseudo packets, as they are set up by (2) The old flow will time out normally, as governed by the "tcp closed" flow timeout setting. At timeout, the normal detection, logging and cleanup code will process it. The flagging of a flow making it 'unfindable' in the flow hash is a bit of a hack. The reason for this approach over for example putting the old flow into a forced timeout queue where it could be timed out, is that such a queue could easily become a contention point. The TCP session reuse case can easily be created by an attacker. In case of multiple packet handlers, this could lead to contention on such a flow timeout queue.
11 years ago
/** Don't return this from the flow hash. It has been replaced. */
#define FLOW_TCP_REUSED BIT_U32(2)
/** Flow was inspected against IP-Only sigs in the toserver direction */
#define FLOW_TOSERVER_IPONLY_SET BIT_U32(3)
/** Flow was inspected against IP-Only sigs in the toclient direction */
#define FLOW_TOCLIENT_IPONLY_SET BIT_U32(4)
/** Packet belonging to this flow should not be inspected at all */
#define FLOW_NOPACKET_INSPECTION BIT_U32(5)
/** Packet payloads belonging to this flow should not be inspected */
#define FLOW_NOPAYLOAD_INSPECTION BIT_U32(6)
/** All packets in this flow should be dropped */
#define FLOW_ACTION_DROP BIT_U32(7)
/** Sgh for toserver direction set (even if it's NULL) */
#define FLOW_SGH_TOSERVER BIT_U32(8)
/** Sgh for toclient direction set (even if it's NULL) */
#define FLOW_SGH_TOCLIENT BIT_U32(9)
/** packet to server direction has been logged in drop file (only in IPS mode) */
#define FLOW_TOSERVER_DROP_LOGGED BIT_U32(10)
/** packet to client direction has been logged in drop file (only in IPS mode) */
#define FLOW_TOCLIENT_DROP_LOGGED BIT_U32(11)
/** flow has alerts */
#define FLOW_HAS_ALERTS BIT_U32(12)
/** Pattern matcher alproto detection done */
#define FLOW_TS_PM_ALPROTO_DETECT_DONE BIT_U32(13)
/** Probing parser alproto detection done */
#define FLOW_TS_PP_ALPROTO_DETECT_DONE BIT_U32(14)
/** Expectation alproto detection done */
#define FLOW_TS_PE_ALPROTO_DETECT_DONE BIT_U32(15)
/** Pattern matcher alproto detection done */
#define FLOW_TC_PM_ALPROTO_DETECT_DONE BIT_U32(16)
/** Probing parser alproto detection done */
#define FLOW_TC_PP_ALPROTO_DETECT_DONE BIT_U32(17)
/** Expectation alproto detection done */
#define FLOW_TC_PE_ALPROTO_DETECT_DONE BIT_U32(18)
#define FLOW_TIMEOUT_REASSEMBLY_DONE BIT_U32(19)
/** flow is ipv4 */
#define FLOW_IPV4 BIT_U32(20)
/** flow is ipv6 */
#define FLOW_IPV6 BIT_U32(21)
#define FLOW_PROTO_DETECT_TS_DONE BIT_U32(22)
#define FLOW_PROTO_DETECT_TC_DONE BIT_U32(23)
/** Indicate that alproto detection for flow should be done again */
#define FLOW_CHANGE_PROTO BIT_U32(24)
#define FLOW_WRONG_THREAD BIT_U32(25)
/** Protocol detection told us flow is picked up in wrong direction (midstream) */
#define FLOW_DIR_REVERSED BIT_U32(26)
/** Indicate that the flow did trigger an expectation creation */
#define FLOW_HAS_EXPECTATION BIT_U32(27)
/** All packets in this flow should be passed */
#define FLOW_ACTION_PASS BIT_U32(28)
/* File flags */
#define FLOWFILE_INIT 0
/** no magic on files in this flow */
#define FLOWFILE_NO_MAGIC_TS BIT_U16(0)
#define FLOWFILE_NO_MAGIC_TC BIT_U16(1)
/** even if the flow has files, don't store 'm */
#define FLOWFILE_NO_STORE_TS BIT_U16(2)
#define FLOWFILE_NO_STORE_TC BIT_U16(3)
/** no md5 on files in this flow */
#define FLOWFILE_NO_MD5_TS BIT_U16(4)
#define FLOWFILE_NO_MD5_TC BIT_U16(5)
/** no sha1 on files in this flow */
#define FLOWFILE_NO_SHA1_TS BIT_U16(6)
#define FLOWFILE_NO_SHA1_TC BIT_U16(7)
/** no sha256 on files in this flow */
#define FLOWFILE_NO_SHA256_TS BIT_U16(8)
#define FLOWFILE_NO_SHA256_TC BIT_U16(9)
/** no size tracking of files in this flow */
#define FLOWFILE_NO_SIZE_TS BIT_U16(10)
#define FLOWFILE_NO_SIZE_TC BIT_U16(11)
#define FLOWFILE_NONE_TS (FLOWFILE_NO_MAGIC_TS | \
FLOWFILE_NO_STORE_TS | \
FLOWFILE_NO_MD5_TS | \
FLOWFILE_NO_SHA1_TS | \
FLOWFILE_NO_SHA256_TS| \
FLOWFILE_NO_SIZE_TS)
#define FLOWFILE_NONE_TC (FLOWFILE_NO_MAGIC_TC | \
FLOWFILE_NO_STORE_TC | \
FLOWFILE_NO_MD5_TC | \
FLOWFILE_NO_SHA1_TC | \
FLOWFILE_NO_SHA256_TC| \
FLOWFILE_NO_SIZE_TC)
#define FLOWFILE_NONE (FLOWFILE_NONE_TS|FLOWFILE_NONE_TC)
#define FLOW_IS_IPV4(f) \
(((f)->flags & FLOW_IPV4) == FLOW_IPV4)
#define FLOW_IS_IPV6(f) \
(((f)->flags & FLOW_IPV6) == FLOW_IPV6)
#define FLOW_GET_SP(f) \
((f)->flags & FLOW_DIR_REVERSED) ? (f)->dp : (f)->sp;
#define FLOW_GET_DP(f) \
((f)->flags & FLOW_DIR_REVERSED) ? (f)->sp : (f)->dp;
#define FLOW_COPY_IPV4_ADDR_TO_PACKET(fa, pa) do { \
(pa)->family = AF_INET; \
(pa)->addr_data32[0] = (fa)->addr_data32[0]; \
} while (0)
#define FLOW_COPY_IPV6_ADDR_TO_PACKET(fa, pa) do { \
(pa)->family = AF_INET6; \
(pa)->addr_data32[0] = (fa)->addr_data32[0]; \
(pa)->addr_data32[1] = (fa)->addr_data32[1]; \
(pa)->addr_data32[2] = (fa)->addr_data32[2]; \
(pa)->addr_data32[3] = (fa)->addr_data32[3]; \
} while (0)
/* Set the IPv4 addressesinto the Addrs of the Packet.
* Make sure p->ip4h is initialized and validated.
*
* We set the rest of the struct to 0 so we can
* prevent using memset. */
#define FLOW_SET_IPV4_SRC_ADDR_FROM_PACKET(p, a) do { \
(a)->addr_data32[0] = (uint32_t)(p)->ip4h->s_ip_src.s_addr; \
(a)->addr_data32[1] = 0; \
(a)->addr_data32[2] = 0; \
(a)->addr_data32[3] = 0; \
} while (0)
#define FLOW_SET_IPV4_DST_ADDR_FROM_PACKET(p, a) do { \
(a)->addr_data32[0] = (uint32_t)(p)->ip4h->s_ip_dst.s_addr; \
(a)->addr_data32[1] = 0; \
(a)->addr_data32[2] = 0; \
(a)->addr_data32[3] = 0; \
} while (0)
/* clear the address structure by setting all fields to 0 */
#define FLOW_CLEAR_ADDR(a) do { \
(a)->addr_data32[0] = 0; \
(a)->addr_data32[1] = 0; \
(a)->addr_data32[2] = 0; \
(a)->addr_data32[3] = 0; \
} while (0)
/* Set the IPv6 addressesinto the Addrs of the Packet.
* Make sure p->ip6h is initialized and validated. */
#define FLOW_SET_IPV6_SRC_ADDR_FROM_PACKET(p, a) do { \
(a)->addr_data32[0] = (p)->ip6h->s_ip6_src[0]; \
(a)->addr_data32[1] = (p)->ip6h->s_ip6_src[1]; \
(a)->addr_data32[2] = (p)->ip6h->s_ip6_src[2]; \
(a)->addr_data32[3] = (p)->ip6h->s_ip6_src[3]; \
} while (0)
#define FLOW_SET_IPV6_DST_ADDR_FROM_PACKET(p, a) do { \
(a)->addr_data32[0] = (p)->ip6h->s_ip6_dst[0]; \
(a)->addr_data32[1] = (p)->ip6h->s_ip6_dst[1]; \
(a)->addr_data32[2] = (p)->ip6h->s_ip6_dst[2]; \
(a)->addr_data32[3] = (p)->ip6h->s_ip6_dst[3]; \
} while (0)
/* pkt flow flags */
#define FLOW_PKT_TOSERVER 0x01
#define FLOW_PKT_TOCLIENT 0x02
#define FLOW_PKT_ESTABLISHED 0x04
#define FLOW_PKT_TOSERVER_IPONLY_SET 0x08
#define FLOW_PKT_TOCLIENT_IPONLY_SET 0x10
#define FLOW_PKT_TOSERVER_FIRST 0x20
#define FLOW_PKT_TOCLIENT_FIRST 0x40
/** last pseudo packet in the flow. Can be used to trigger final clean,
* logging, etc. */
#define FLOW_PKT_LAST_PSEUDO 0x80
#define FLOW_END_FLAG_STATE_NEW 0x01
#define FLOW_END_FLAG_STATE_ESTABLISHED 0x02
#define FLOW_END_FLAG_STATE_CLOSED 0x04
#define FLOW_END_FLAG_EMERGENCY 0x08
#define FLOW_END_FLAG_TIMEOUT 0x10
#define FLOW_END_FLAG_FORCED 0x20
#define FLOW_END_FLAG_SHUTDOWN 0x40
#define FLOW_END_FLAG_STATE_BYPASSED 0x80
/** Mutex or RWLocks for the flow. */
//#define FLOWLOCK_RWLOCK
#define FLOWLOCK_MUTEX
#ifdef FLOWLOCK_RWLOCK
#ifdef FLOWLOCK_MUTEX
#error Cannot enable both FLOWLOCK_RWLOCK and FLOWLOCK_MUTEX
#endif
#endif
#ifdef FLOWLOCK_RWLOCK
#define FLOWLOCK_INIT(fb) SCRWLockInit(&(fb)->r, NULL)
#define FLOWLOCK_DESTROY(fb) SCRWLockDestroy(&(fb)->r)
#define FLOWLOCK_RDLOCK(fb) SCRWLockRDLock(&(fb)->r)
#define FLOWLOCK_WRLOCK(fb) SCRWLockWRLock(&(fb)->r)
#define FLOWLOCK_TRYRDLOCK(fb) SCRWLockTryRDLock(&(fb)->r)
#define FLOWLOCK_TRYWRLOCK(fb) SCRWLockTryWRLock(&(fb)->r)
#define FLOWLOCK_UNLOCK(fb) SCRWLockUnlock(&(fb)->r)
#elif defined FLOWLOCK_MUTEX
#define FLOWLOCK_INIT(fb) SCMutexInit(&(fb)->m, NULL)
#define FLOWLOCK_DESTROY(fb) SCMutexDestroy(&(fb)->m)
#define FLOWLOCK_RDLOCK(fb) SCMutexLock(&(fb)->m)
#define FLOWLOCK_WRLOCK(fb) SCMutexLock(&(fb)->m)
#define FLOWLOCK_TRYRDLOCK(fb) SCMutexTrylock(&(fb)->m)
#define FLOWLOCK_TRYWRLOCK(fb) SCMutexTrylock(&(fb)->m)
#define FLOWLOCK_UNLOCK(fb) SCMutexUnlock(&(fb)->m)
#else
#error Enable FLOWLOCK_RWLOCK or FLOWLOCK_MUTEX
#endif
App layer API rewritten. The main files in question are: app-layer.[ch], app-layer-detect-proto.[ch] and app-layer-parser.[ch]. Things addressed in this commit: - Brings out a proper separation between protocol detection phase and the parser phase. - The dns app layer now is registered such that we don't use "dnstcp" and "dnsudp" in the rules. A user who previously wrote a rule like this - "alert dnstcp....." or "alert dnsudp....." would now have to use, alert dns (ipproto:tcp;) or alert udp (app-layer-protocol:dns;) or alert ip (ipproto:udp; app-layer-protocol:dns;) The same rules extend to other another such protocol, dcerpc. - The app layer parser api now takes in the ipproto while registering callbacks. - The app inspection/detection engine also takes an ipproto. - All app layer parser functions now take direction as STREAM_TOSERVER or STREAM_TOCLIENT, as opposed to 0 or 1, which was taken by some of the functions. - FlowInitialize() and FlowRecycle() now resets proto to 0. This is needed by unittests, which would try to clean the flow, and that would call the api, AppLayerParserCleanupParserState(), which would try to clean the app state, but the app layer now needs an ipproto to figure out which api to internally call to clean the state, and if the ipproto is 0, it would return without trying to clean the state. - A lot of unittests are now updated where if they are using a flow and they need to use the app layer, we would set a flow ipproto. - The "app-layer" section in the yaml conf has also been updated as well.
12 years ago
#define FLOW_IS_PM_DONE(f, dir) (((dir) & STREAM_TOSERVER) ? ((f)->flags & FLOW_TS_PM_ALPROTO_DETECT_DONE) : ((f)->flags & FLOW_TC_PM_ALPROTO_DETECT_DONE))
#define FLOW_IS_PP_DONE(f, dir) (((dir) & STREAM_TOSERVER) ? ((f)->flags & FLOW_TS_PP_ALPROTO_DETECT_DONE) : ((f)->flags & FLOW_TC_PP_ALPROTO_DETECT_DONE))
#define FLOW_IS_PE_DONE(f, dir) (((dir) & STREAM_TOSERVER) ? ((f)->flags & FLOW_TS_PE_ALPROTO_DETECT_DONE) : ((f)->flags & FLOW_TC_PE_ALPROTO_DETECT_DONE))
App layer API rewritten. The main files in question are: app-layer.[ch], app-layer-detect-proto.[ch] and app-layer-parser.[ch]. Things addressed in this commit: - Brings out a proper separation between protocol detection phase and the parser phase. - The dns app layer now is registered such that we don't use "dnstcp" and "dnsudp" in the rules. A user who previously wrote a rule like this - "alert dnstcp....." or "alert dnsudp....." would now have to use, alert dns (ipproto:tcp;) or alert udp (app-layer-protocol:dns;) or alert ip (ipproto:udp; app-layer-protocol:dns;) The same rules extend to other another such protocol, dcerpc. - The app layer parser api now takes in the ipproto while registering callbacks. - The app inspection/detection engine also takes an ipproto. - All app layer parser functions now take direction as STREAM_TOSERVER or STREAM_TOCLIENT, as opposed to 0 or 1, which was taken by some of the functions. - FlowInitialize() and FlowRecycle() now resets proto to 0. This is needed by unittests, which would try to clean the flow, and that would call the api, AppLayerParserCleanupParserState(), which would try to clean the app state, but the app layer now needs an ipproto to figure out which api to internally call to clean the state, and if the ipproto is 0, it would return without trying to clean the state. - A lot of unittests are now updated where if they are using a flow and they need to use the app layer, we would set a flow ipproto. - The "app-layer" section in the yaml conf has also been updated as well.
12 years ago
#define FLOW_SET_PM_DONE(f, dir) (((dir) & STREAM_TOSERVER) ? ((f)->flags |= FLOW_TS_PM_ALPROTO_DETECT_DONE) : ((f)->flags |= FLOW_TC_PM_ALPROTO_DETECT_DONE))
#define FLOW_SET_PP_DONE(f, dir) (((dir) & STREAM_TOSERVER) ? ((f)->flags |= FLOW_TS_PP_ALPROTO_DETECT_DONE) : ((f)->flags |= FLOW_TC_PP_ALPROTO_DETECT_DONE))
#define FLOW_SET_PE_DONE(f, dir) (((dir) & STREAM_TOSERVER) ? ((f)->flags |= FLOW_TS_PE_ALPROTO_DETECT_DONE) : ((f)->flags |= FLOW_TC_PE_ALPROTO_DETECT_DONE))
App layer API rewritten. The main files in question are: app-layer.[ch], app-layer-detect-proto.[ch] and app-layer-parser.[ch]. Things addressed in this commit: - Brings out a proper separation between protocol detection phase and the parser phase. - The dns app layer now is registered such that we don't use "dnstcp" and "dnsudp" in the rules. A user who previously wrote a rule like this - "alert dnstcp....." or "alert dnsudp....." would now have to use, alert dns (ipproto:tcp;) or alert udp (app-layer-protocol:dns;) or alert ip (ipproto:udp; app-layer-protocol:dns;) The same rules extend to other another such protocol, dcerpc. - The app layer parser api now takes in the ipproto while registering callbacks. - The app inspection/detection engine also takes an ipproto. - All app layer parser functions now take direction as STREAM_TOSERVER or STREAM_TOCLIENT, as opposed to 0 or 1, which was taken by some of the functions. - FlowInitialize() and FlowRecycle() now resets proto to 0. This is needed by unittests, which would try to clean the flow, and that would call the api, AppLayerParserCleanupParserState(), which would try to clean the app state, but the app layer now needs an ipproto to figure out which api to internally call to clean the state, and if the ipproto is 0, it would return without trying to clean the state. - A lot of unittests are now updated where if they are using a flow and they need to use the app layer, we would set a flow ipproto. - The "app-layer" section in the yaml conf has also been updated as well.
12 years ago
#define FLOW_RESET_PM_DONE(f, dir) (((dir) & STREAM_TOSERVER) ? ((f)->flags &= ~FLOW_TS_PM_ALPROTO_DETECT_DONE) : ((f)->flags &= ~FLOW_TC_PM_ALPROTO_DETECT_DONE))
#define FLOW_RESET_PP_DONE(f, dir) (((dir) & STREAM_TOSERVER) ? ((f)->flags &= ~FLOW_TS_PP_ALPROTO_DETECT_DONE) : ((f)->flags &= ~FLOW_TC_PP_ALPROTO_DETECT_DONE))
#define FLOW_RESET_PE_DONE(f, dir) (((dir) & STREAM_TOSERVER) ? ((f)->flags &= ~FLOW_TS_PE_ALPROTO_DETECT_DONE) : ((f)->flags &= ~FLOW_TC_PE_ALPROTO_DETECT_DONE))
App layer API rewritten. The main files in question are: app-layer.[ch], app-layer-detect-proto.[ch] and app-layer-parser.[ch]. Things addressed in this commit: - Brings out a proper separation between protocol detection phase and the parser phase. - The dns app layer now is registered such that we don't use "dnstcp" and "dnsudp" in the rules. A user who previously wrote a rule like this - "alert dnstcp....." or "alert dnsudp....." would now have to use, alert dns (ipproto:tcp;) or alert udp (app-layer-protocol:dns;) or alert ip (ipproto:udp; app-layer-protocol:dns;) The same rules extend to other another such protocol, dcerpc. - The app layer parser api now takes in the ipproto while registering callbacks. - The app inspection/detection engine also takes an ipproto. - All app layer parser functions now take direction as STREAM_TOSERVER or STREAM_TOCLIENT, as opposed to 0 or 1, which was taken by some of the functions. - FlowInitialize() and FlowRecycle() now resets proto to 0. This is needed by unittests, which would try to clean the flow, and that would call the api, AppLayerParserCleanupParserState(), which would try to clean the app state, but the app layer now needs an ipproto to figure out which api to internally call to clean the state, and if the ipproto is 0, it would return without trying to clean the state. - A lot of unittests are now updated where if they are using a flow and they need to use the app layer, we would set a flow ipproto. - The "app-layer" section in the yaml conf has also been updated as well.
12 years ago
/* global flow config */
typedef struct FlowCnf_
{
uint32_t hash_rand;
uint32_t hash_size;
uint32_t max_flows;
uint32_t prealloc;
uint32_t timeout_new;
uint32_t timeout_est;
uint32_t emerg_timeout_new;
uint32_t emerg_timeout_est;
uint32_t emergency_recovery;
enum ExceptionPolicy memcap_policy;
SC_ATOMIC_DECLARE(uint64_t, memcap);
} FlowConfig;
16 years ago
/* Hash key for the flow hash */
typedef struct FlowKey_
{
Address src, dst;
Port sp, dp;
uint8_t proto;
uint8_t recursion_level;
uint16_t vlan_id[2];
} FlowKey;
typedef struct FlowAddress_ {
union {
uint32_t address_un_data32[4]; /* type-specific field */
uint16_t address_un_data16[8]; /* type-specific field */
uint8_t address_un_data8[16]; /* type-specific field */
} address;
} FlowAddress;
#define addr_data32 address.address_un_data32
#define addr_data16 address.address_un_data16
#define addr_data8 address.address_un_data8
typedef unsigned short FlowRefCount;
typedef unsigned short FlowStateType;
/** Local Thread ID */
typedef uint16_t FlowThreadId;
/**
* \brief Flow data structure.
*
* The flow is a global data structure that is created for new packets of a
* flow and then looked up for the following packets of a flow.
*
* Locking
*
* The flow is updated/used by multiple packets at the same time. This is why
* there is a flow-mutex. It's a mutex and not a spinlock because some
* operations on the flow can be quite expensive, thus spinning would be
* too expensive.
*
* The flow "header" (addresses, ports, proto, recursion level) are static
* after the initialization and remain read-only throughout the entire live
* of a flow. This is why we can access those without protection of the lock.
*/
typedef struct Flow_
{
/* flow "header", used for hashing and flow lookup. Static after init,
* so safe to look at without lock */
FlowAddress src, dst;
union {
Port sp; /**< tcp/udp source port */
struct {
uint8_t type; /**< icmp type */
uint8_t code; /**< icmp code */
} icmp_s;
struct {
uint32_t spi; /**< esp spi */
} esp;
};
union {
Port dp; /**< tcp/udp destination port */
struct {
uint8_t type; /**< icmp type */
uint8_t code; /**< icmp code */
} icmp_d;
};
uint8_t proto;
uint8_t recursion_level;
uint16_t vlan_id[2];
flow: redesign of flow timeout handling Goals: - reduce locking - take advantage of 'hot' caches - better locality Locking reduction New flow spare pool. The global pool is implmented as a list of blocks, where each block has a 100 spare flows. Worker threads fetch a block at a time, storing the block in the local thread storage. Flow Recycler now returns flows to the pool is blocks as well. Flow Recycler fetches all flows to be processed in one step instead of one at a time. Cache 'hot'ness Worker threads now check the timeout of flows they evaluate during lookup. The worker will have to read the flow into cache anyway, so the added overhead of checking the timeout value is minimal. When a flow is considered timed out, one of 2 things happens: - if the flow is 'owned' by the thread it is handled locally. Handling means checking if the flow needs 'timeout' work. - otherwise, the flow is added to a special 'evicted' list in the flow bucket where it will be picked up by the flow manager. Flow Manager timing By default the flow manager now tries to do passes of the flow hash in smaller steps, where the goal is to do full pass in 8 x the lowest timeout value it has to enforce. So if the lowest timeout value is 30s, a full pass will take 4 minutes. The goal here is to reduce locking overhead and not get in the way of the workers. In emergency mode each pass is full, and lower timeouts are used. Timing of the flow manager is also no longer relying on pthread condition variables, as these generally cause waking up much quicker than the desired timout. Instead a simple (u)sleep loop is used. Both changes reduce the number of hash passes a lot. Emergency behavior In emergency mode there a number of changes to the workers. In this scenario the flow memcap is fully used up and it is unavoidable that some flows won't be tracked. 1. flow spare pool fetches are reduced to once a second. This avoids locking overhead, while the chance of success was very low. 2. getting an active flow directly from the hash skips flows that had very recent activity to avoid the scenario where all flows get only into the NEW state before getting reused. Rather allow some to have a chance of completing. 3. TCP packets that are not SYN packets will not get a used flow, unless stream.midstream is enabled. The goal here is again to avoid evicting active flows unnecessarily. Better Localily Flow Manager injects flows into the worker threads now, instead of one or two packets. Advantage of this is that the worker threads can get packets from their local packet pools, avoiding constant overhead of packets returning to 'foreign' pools. Counters A lot of flow counters have been added and some have been renamed. Overall the worker threads increment 'flow.wrk.*' counters, while the flow manager increments 'flow.mgr.*'. Additionally, none of the counters are snapshots anymore, they all increment over time. The flow.memuse and flow.spare counters are exceptions. Misc FlowQueue has been split into a FlowQueuePrivate (unlocked) and FlowQueue. Flow no longer has 'prev' pointers and used a unified 'next' pointer for both hash and queue use.
6 years ago
/** how many references exist to this flow *right now*
*
* On receiving a packet the counter is incremented while the flow
* bucked is locked, which is also the case on timeout pruning.
*/
FlowRefCount use_cnt;
uint8_t vlan_idx;
flow: redesign of flow timeout handling Goals: - reduce locking - take advantage of 'hot' caches - better locality Locking reduction New flow spare pool. The global pool is implmented as a list of blocks, where each block has a 100 spare flows. Worker threads fetch a block at a time, storing the block in the local thread storage. Flow Recycler now returns flows to the pool is blocks as well. Flow Recycler fetches all flows to be processed in one step instead of one at a time. Cache 'hot'ness Worker threads now check the timeout of flows they evaluate during lookup. The worker will have to read the flow into cache anyway, so the added overhead of checking the timeout value is minimal. When a flow is considered timed out, one of 2 things happens: - if the flow is 'owned' by the thread it is handled locally. Handling means checking if the flow needs 'timeout' work. - otherwise, the flow is added to a special 'evicted' list in the flow bucket where it will be picked up by the flow manager. Flow Manager timing By default the flow manager now tries to do passes of the flow hash in smaller steps, where the goal is to do full pass in 8 x the lowest timeout value it has to enforce. So if the lowest timeout value is 30s, a full pass will take 4 minutes. The goal here is to reduce locking overhead and not get in the way of the workers. In emergency mode each pass is full, and lower timeouts are used. Timing of the flow manager is also no longer relying on pthread condition variables, as these generally cause waking up much quicker than the desired timout. Instead a simple (u)sleep loop is used. Both changes reduce the number of hash passes a lot. Emergency behavior In emergency mode there a number of changes to the workers. In this scenario the flow memcap is fully used up and it is unavoidable that some flows won't be tracked. 1. flow spare pool fetches are reduced to once a second. This avoids locking overhead, while the chance of success was very low. 2. getting an active flow directly from the hash skips flows that had very recent activity to avoid the scenario where all flows get only into the NEW state before getting reused. Rather allow some to have a chance of completing. 3. TCP packets that are not SYN packets will not get a used flow, unless stream.midstream is enabled. The goal here is again to avoid evicting active flows unnecessarily. Better Localily Flow Manager injects flows into the worker threads now, instead of one or two packets. Advantage of this is that the worker threads can get packets from their local packet pools, avoiding constant overhead of packets returning to 'foreign' pools. Counters A lot of flow counters have been added and some have been renamed. Overall the worker threads increment 'flow.wrk.*' counters, while the flow manager increments 'flow.mgr.*'. Additionally, none of the counters are snapshots anymore, they all increment over time. The flow.memuse and flow.spare counters are exceptions. Misc FlowQueue has been split into a FlowQueuePrivate (unlocked) and FlowQueue. Flow no longer has 'prev' pointers and used a unified 'next' pointer for both hash and queue use.
6 years ago
/* track toserver/toclient flow timeout needs */
union {
struct {
uint8_t ffr_ts:4;
uint8_t ffr_tc:4;
};
uint8_t ffr;
};
/** timestamp in seconds of the moment this flow will timeout
* according to the timeout policy. Does *not* take emergency
* mode into account. */
uint32_t timeout_at;
/** Thread ID for the stream/detect portion of this flow */
FlowThreadId thread_id[2];
struct Flow_ *next; /* (hash) list next */
/** Incoming interface */
struct LiveDevice_ *livedev;
/** flow hash - the flow hash before hash table size mod. */
uint32_t flow_hash;
/* time stamp of last update (last packet). Set/updated under the
* flow and flow hash row locks, safe to read under either the
* flow lock or flow hash row lock. */
struct timeval lastts;
/* end of flow "header" */
flow: redesign of flow timeout handling Goals: - reduce locking - take advantage of 'hot' caches - better locality Locking reduction New flow spare pool. The global pool is implmented as a list of blocks, where each block has a 100 spare flows. Worker threads fetch a block at a time, storing the block in the local thread storage. Flow Recycler now returns flows to the pool is blocks as well. Flow Recycler fetches all flows to be processed in one step instead of one at a time. Cache 'hot'ness Worker threads now check the timeout of flows they evaluate during lookup. The worker will have to read the flow into cache anyway, so the added overhead of checking the timeout value is minimal. When a flow is considered timed out, one of 2 things happens: - if the flow is 'owned' by the thread it is handled locally. Handling means checking if the flow needs 'timeout' work. - otherwise, the flow is added to a special 'evicted' list in the flow bucket where it will be picked up by the flow manager. Flow Manager timing By default the flow manager now tries to do passes of the flow hash in smaller steps, where the goal is to do full pass in 8 x the lowest timeout value it has to enforce. So if the lowest timeout value is 30s, a full pass will take 4 minutes. The goal here is to reduce locking overhead and not get in the way of the workers. In emergency mode each pass is full, and lower timeouts are used. Timing of the flow manager is also no longer relying on pthread condition variables, as these generally cause waking up much quicker than the desired timout. Instead a simple (u)sleep loop is used. Both changes reduce the number of hash passes a lot. Emergency behavior In emergency mode there a number of changes to the workers. In this scenario the flow memcap is fully used up and it is unavoidable that some flows won't be tracked. 1. flow spare pool fetches are reduced to once a second. This avoids locking overhead, while the chance of success was very low. 2. getting an active flow directly from the hash skips flows that had very recent activity to avoid the scenario where all flows get only into the NEW state before getting reused. Rather allow some to have a chance of completing. 3. TCP packets that are not SYN packets will not get a used flow, unless stream.midstream is enabled. The goal here is again to avoid evicting active flows unnecessarily. Better Localily Flow Manager injects flows into the worker threads now, instead of one or two packets. Advantage of this is that the worker threads can get packets from their local packet pools, avoiding constant overhead of packets returning to 'foreign' pools. Counters A lot of flow counters have been added and some have been renamed. Overall the worker threads increment 'flow.wrk.*' counters, while the flow manager increments 'flow.mgr.*'. Additionally, none of the counters are snapshots anymore, they all increment over time. The flow.memuse and flow.spare counters are exceptions. Misc FlowQueue has been split into a FlowQueuePrivate (unlocked) and FlowQueue. Flow no longer has 'prev' pointers and used a unified 'next' pointer for both hash and queue use.
6 years ago
/** timeout policy value in seconds to add to the lastts.tv_sec
* when a packet has been received. */
uint32_t timeout_policy;
flow: redesign of flow timeout handling Goals: - reduce locking - take advantage of 'hot' caches - better locality Locking reduction New flow spare pool. The global pool is implmented as a list of blocks, where each block has a 100 spare flows. Worker threads fetch a block at a time, storing the block in the local thread storage. Flow Recycler now returns flows to the pool is blocks as well. Flow Recycler fetches all flows to be processed in one step instead of one at a time. Cache 'hot'ness Worker threads now check the timeout of flows they evaluate during lookup. The worker will have to read the flow into cache anyway, so the added overhead of checking the timeout value is minimal. When a flow is considered timed out, one of 2 things happens: - if the flow is 'owned' by the thread it is handled locally. Handling means checking if the flow needs 'timeout' work. - otherwise, the flow is added to a special 'evicted' list in the flow bucket where it will be picked up by the flow manager. Flow Manager timing By default the flow manager now tries to do passes of the flow hash in smaller steps, where the goal is to do full pass in 8 x the lowest timeout value it has to enforce. So if the lowest timeout value is 30s, a full pass will take 4 minutes. The goal here is to reduce locking overhead and not get in the way of the workers. In emergency mode each pass is full, and lower timeouts are used. Timing of the flow manager is also no longer relying on pthread condition variables, as these generally cause waking up much quicker than the desired timout. Instead a simple (u)sleep loop is used. Both changes reduce the number of hash passes a lot. Emergency behavior In emergency mode there a number of changes to the workers. In this scenario the flow memcap is fully used up and it is unavoidable that some flows won't be tracked. 1. flow spare pool fetches are reduced to once a second. This avoids locking overhead, while the chance of success was very low. 2. getting an active flow directly from the hash skips flows that had very recent activity to avoid the scenario where all flows get only into the NEW state before getting reused. Rather allow some to have a chance of completing. 3. TCP packets that are not SYN packets will not get a used flow, unless stream.midstream is enabled. The goal here is again to avoid evicting active flows unnecessarily. Better Localily Flow Manager injects flows into the worker threads now, instead of one or two packets. Advantage of this is that the worker threads can get packets from their local packet pools, avoiding constant overhead of packets returning to 'foreign' pools. Counters A lot of flow counters have been added and some have been renamed. Overall the worker threads increment 'flow.wrk.*' counters, while the flow manager increments 'flow.mgr.*'. Additionally, none of the counters are snapshots anymore, they all increment over time. The flow.memuse and flow.spare counters are exceptions. Misc FlowQueue has been split into a FlowQueuePrivate (unlocked) and FlowQueue. Flow no longer has 'prev' pointers and used a unified 'next' pointer for both hash and queue use.
6 years ago
FlowStateType flow_state;
/** flow tenant id, used to setup flow timeout and stream pseudo
* packets with the correct tenant id set */
uint32_t tenant_id;
App layer API rewritten. The main files in question are: app-layer.[ch], app-layer-detect-proto.[ch] and app-layer-parser.[ch]. Things addressed in this commit: - Brings out a proper separation between protocol detection phase and the parser phase. - The dns app layer now is registered such that we don't use "dnstcp" and "dnsudp" in the rules. A user who previously wrote a rule like this - "alert dnstcp....." or "alert dnsudp....." would now have to use, alert dns (ipproto:tcp;) or alert udp (app-layer-protocol:dns;) or alert ip (ipproto:udp; app-layer-protocol:dns;) The same rules extend to other another such protocol, dcerpc. - The app layer parser api now takes in the ipproto while registering callbacks. - The app inspection/detection engine also takes an ipproto. - All app layer parser functions now take direction as STREAM_TOSERVER or STREAM_TOCLIENT, as opposed to 0 or 1, which was taken by some of the functions. - FlowInitialize() and FlowRecycle() now resets proto to 0. This is needed by unittests, which would try to clean the flow, and that would call the api, AppLayerParserCleanupParserState(), which would try to clean the app state, but the app layer now needs an ipproto to figure out which api to internally call to clean the state, and if the ipproto is 0, it would return without trying to clean the state. - A lot of unittests are now updated where if they are using a flow and they need to use the app layer, we would set a flow ipproto. - The "app-layer" section in the yaml conf has also been updated as well.
12 years ago
uint32_t probing_parser_toserver_alproto_masks;
uint32_t probing_parser_toclient_alproto_masks;
uint32_t flags; /**< generic flags */
uint16_t file_flags; /**< file tracking/extraction flags */
/** destination port to be used in protocol detection. This is meant
* for use with STARTTLS and HTTP CONNECT detection */
uint16_t protodetect_dp; /**< 0 if not used */
/* Parent flow id for protocol like ftp */
int64_t parent_id;
#ifdef FLOWLOCK_RWLOCK
SCRWLock r;
#elif defined FLOWLOCK_MUTEX
SCMutex m;
#else
#error Enable FLOWLOCK_RWLOCK or FLOWLOCK_MUTEX
#endif
/** protocol specific data pointer, e.g. for TcpSession */
void *protoctx;
/** mapping to Flow's protocol specific protocols for timeouts
and state and free functions. */
uint8_t protomap;
uint8_t flow_end_flags;
/* coccinelle: Flow:flow_end_flags:FLOW_END_FLAG_ */
AppProto alproto; /**< \brief application level protocol */
AppProto alproto_ts;
AppProto alproto_tc;
/** original application level protocol. Used to indicate the previous
protocol when changing to another protocol , e.g. with STARTTLS. */
AppProto alproto_orig;
/** expected app protocol: used in protocol change/upgrade like in
* STARTTLS. */
AppProto alproto_expect;
/** detection engine ctx version used to inspect this flow. Set at initial
* inspection. If it doesn't match the currently in use de_ctx, the
* stored sgh ptrs are reset. */
uint32_t de_ctx_version;
/** ttl tracking */
uint8_t min_ttl_toserver;
uint8_t max_ttl_toserver;
uint8_t min_ttl_toclient;
uint8_t max_ttl_toclient;
/** application level storage ptrs.
*
*/
AppLayerParserState *alparser; /**< parser internal state */
void *alstate; /**< application layer state */
/** toclient sgh for this flow. Only use when FLOW_SGH_TOCLIENT flow flag
* has been set. */
const struct SigGroupHead_ *sgh_toclient;
/** toserver sgh for this flow. Only use when FLOW_SGH_TOSERVER flow flag
* has been set. */
const struct SigGroupHead_ *sgh_toserver;
/* pointer to the var list */
GenericVar *flowvar;
struct FlowBucket_ *fb;
struct timeval startts;
uint32_t todstpktcnt;
uint32_t tosrcpktcnt;
uint64_t todstbytecnt;
uint64_t tosrcbytecnt;
} Flow;
flow-manager: optimize hash walking Until now the flow manager would walk the entire flow hash table on an interval. It would thus touch all flows, leading to a lot of memory and cache pressure. In scenario's where the number of tracked flows run into the hundreds on thousands, and the memory used can run into many hundreds of megabytes or even gigabytes, this would lead to serious performance degradation. This patch introduces a new approach. A timestamp per flow bucket (hash row) is maintained by the flow manager. It holds the timestamp of the earliest possible timeout of a flow in the list. The hash walk skips rows with timestamps beyond the current time. As the timestamp depends on the flows in the hash row's list, and on the 'state' of each flow in the list, any addition of a flow or changing of a flow's state invalidates the timestamp. The flow manager then has to walk the list again to set a new timestamp. A utility function FlowUpdateState is introduced to change Flow states, taking care of the bucket timestamp invalidation while at it. Empty flow buckets use a special value so that we don't have to take the flow bucket lock to find out the bucket is empty. This patch also adds more performance counters: flow_mgr.flows_checked | Total | 929 flow_mgr.flows_notimeout | Total | 391 flow_mgr.flows_timeout | Total | 538 flow_mgr.flows_removed | Total | 277 flow_mgr.flows_timeout_inuse | Total | 261 flow_mgr.rows_checked | Total | 1000000 flow_mgr.rows_skipped | Total | 998835 flow_mgr.rows_empty | Total | 290 flow_mgr.rows_maxlen | Total | 2 flow_mgr.flows_checked: number of flows checked for timeout in the last pass flow_mgr.flows_notimeout: number of flows out of flow_mgr.flows_checked that didn't time out flow_mgr.flows_timeout: number of out of flow_mgr.flows_checked that did reach the time out flow_mgr.flows_removed: number of flows out of flow_mgr.flows_timeout that were really removed flow_mgr.flows_timeout_inuse: number of flows out of flow_mgr.flows_timeout that were still in use or needed work flow_mgr.rows_checked: hash table rows checked flow_mgr.rows_skipped: hash table rows skipped because non of the flows would time out anyway The counters below are only relating to rows that were not skipped. flow_mgr.rows_empty: empty hash rows flow_mgr.rows_maxlen: max number of flows per hash row. Best to keep low, so increase hash-size if needed. flow_mgr.rows_busy: row skipped because it was locked by another thread
9 years ago
enum FlowState {
FLOW_STATE_NEW = 0,
FLOW_STATE_ESTABLISHED,
FLOW_STATE_CLOSED,
FLOW_STATE_LOCAL_BYPASSED,
#ifdef CAPTURE_OFFLOAD
FLOW_STATE_CAPTURE_BYPASSED,
#endif
};
#ifdef CAPTURE_OFFLOAD
#define FLOW_STATE_SIZE 5
#else
#define FLOW_STATE_SIZE 4
#endif
typedef struct FlowProtoTimeout_ {
uint32_t new_timeout;
uint32_t est_timeout;
uint32_t closed_timeout;
uint32_t bypassed_timeout;
} FlowProtoTimeout;
typedef struct FlowProtoFreeFunc_ {
void (*Freefunc)(void *);
} FlowProtoFreeFunc;
typedef struct FlowBypassInfo_ {
bool (* BypassUpdate)(Flow *f, void *data, time_t tsec);
void (* BypassFree)(void *data);
void *bypass_data;
uint64_t tosrcpktcnt;
uint64_t tosrcbytecnt;
uint64_t todstpktcnt;
uint64_t todstbytecnt;
} FlowBypassInfo;
flow: redesign of flow timeout handling Goals: - reduce locking - take advantage of 'hot' caches - better locality Locking reduction New flow spare pool. The global pool is implmented as a list of blocks, where each block has a 100 spare flows. Worker threads fetch a block at a time, storing the block in the local thread storage. Flow Recycler now returns flows to the pool is blocks as well. Flow Recycler fetches all flows to be processed in one step instead of one at a time. Cache 'hot'ness Worker threads now check the timeout of flows they evaluate during lookup. The worker will have to read the flow into cache anyway, so the added overhead of checking the timeout value is minimal. When a flow is considered timed out, one of 2 things happens: - if the flow is 'owned' by the thread it is handled locally. Handling means checking if the flow needs 'timeout' work. - otherwise, the flow is added to a special 'evicted' list in the flow bucket where it will be picked up by the flow manager. Flow Manager timing By default the flow manager now tries to do passes of the flow hash in smaller steps, where the goal is to do full pass in 8 x the lowest timeout value it has to enforce. So if the lowest timeout value is 30s, a full pass will take 4 minutes. The goal here is to reduce locking overhead and not get in the way of the workers. In emergency mode each pass is full, and lower timeouts are used. Timing of the flow manager is also no longer relying on pthread condition variables, as these generally cause waking up much quicker than the desired timout. Instead a simple (u)sleep loop is used. Both changes reduce the number of hash passes a lot. Emergency behavior In emergency mode there a number of changes to the workers. In this scenario the flow memcap is fully used up and it is unavoidable that some flows won't be tracked. 1. flow spare pool fetches are reduced to once a second. This avoids locking overhead, while the chance of success was very low. 2. getting an active flow directly from the hash skips flows that had very recent activity to avoid the scenario where all flows get only into the NEW state before getting reused. Rather allow some to have a chance of completing. 3. TCP packets that are not SYN packets will not get a used flow, unless stream.midstream is enabled. The goal here is again to avoid evicting active flows unnecessarily. Better Localily Flow Manager injects flows into the worker threads now, instead of one or two packets. Advantage of this is that the worker threads can get packets from their local packet pools, avoiding constant overhead of packets returning to 'foreign' pools. Counters A lot of flow counters have been added and some have been renamed. Overall the worker threads increment 'flow.wrk.*' counters, while the flow manager increments 'flow.mgr.*'. Additionally, none of the counters are snapshots anymore, they all increment over time. The flow.memuse and flow.spare counters are exceptions. Misc FlowQueue has been split into a FlowQueuePrivate (unlocked) and FlowQueue. Flow no longer has 'prev' pointers and used a unified 'next' pointer for both hash and queue use.
6 years ago
#include "flow-queue.h"
typedef struct FlowLookupStruct_ // TODO name
{
/** thread store of spare queues */
FlowQueuePrivate spare_queue;
DecodeThreadVars *dtv;
FlowQueuePrivate work_queue;
uint32_t emerg_spare_sync_stamp;
} FlowLookupStruct;
/** \brief prepare packet for a life with flow
* Set PKT_WANTS_FLOW flag to incidate workers should do a flow lookup
* and calc the hash value to be used in the lookup and autofp flow
* balancing. */
void FlowSetupPacket(Packet *p);
flow: redesign of flow timeout handling Goals: - reduce locking - take advantage of 'hot' caches - better locality Locking reduction New flow spare pool. The global pool is implmented as a list of blocks, where each block has a 100 spare flows. Worker threads fetch a block at a time, storing the block in the local thread storage. Flow Recycler now returns flows to the pool is blocks as well. Flow Recycler fetches all flows to be processed in one step instead of one at a time. Cache 'hot'ness Worker threads now check the timeout of flows they evaluate during lookup. The worker will have to read the flow into cache anyway, so the added overhead of checking the timeout value is minimal. When a flow is considered timed out, one of 2 things happens: - if the flow is 'owned' by the thread it is handled locally. Handling means checking if the flow needs 'timeout' work. - otherwise, the flow is added to a special 'evicted' list in the flow bucket where it will be picked up by the flow manager. Flow Manager timing By default the flow manager now tries to do passes of the flow hash in smaller steps, where the goal is to do full pass in 8 x the lowest timeout value it has to enforce. So if the lowest timeout value is 30s, a full pass will take 4 minutes. The goal here is to reduce locking overhead and not get in the way of the workers. In emergency mode each pass is full, and lower timeouts are used. Timing of the flow manager is also no longer relying on pthread condition variables, as these generally cause waking up much quicker than the desired timout. Instead a simple (u)sleep loop is used. Both changes reduce the number of hash passes a lot. Emergency behavior In emergency mode there a number of changes to the workers. In this scenario the flow memcap is fully used up and it is unavoidable that some flows won't be tracked. 1. flow spare pool fetches are reduced to once a second. This avoids locking overhead, while the chance of success was very low. 2. getting an active flow directly from the hash skips flows that had very recent activity to avoid the scenario where all flows get only into the NEW state before getting reused. Rather allow some to have a chance of completing. 3. TCP packets that are not SYN packets will not get a used flow, unless stream.midstream is enabled. The goal here is again to avoid evicting active flows unnecessarily. Better Localily Flow Manager injects flows into the worker threads now, instead of one or two packets. Advantage of this is that the worker threads can get packets from their local packet pools, avoiding constant overhead of packets returning to 'foreign' pools. Counters A lot of flow counters have been added and some have been renamed. Overall the worker threads increment 'flow.wrk.*' counters, while the flow manager increments 'flow.mgr.*'. Additionally, none of the counters are snapshots anymore, they all increment over time. The flow.memuse and flow.spare counters are exceptions. Misc FlowQueue has been split into a FlowQueuePrivate (unlocked) and FlowQueue. Flow no longer has 'prev' pointers and used a unified 'next' pointer for both hash and queue use.
6 years ago
void FlowHandlePacket (ThreadVars *, FlowLookupStruct *, Packet *);
void FlowInitConfig(bool);
16 years ago
void FlowPrintQueueInfo (void);
void FlowReset(void);
void FlowShutdown(void);
void FlowSetIPOnlyFlag(Flow *, int);
void FlowSetHasAlertsFlag(Flow *);
int FlowHasAlerts(const Flow *);
void FlowSetChangeProtoFlag(Flow *);
void FlowUnsetChangeProtoFlag(Flow *);
int FlowChangeProto(Flow *);
void FlowSwap(Flow *);
void FlowRegisterTests (void);
int FlowSetProtoTimeout(uint8_t ,uint32_t ,uint32_t ,uint32_t);
int FlowSetProtoEmergencyTimeout(uint8_t ,uint32_t ,uint32_t ,uint32_t);
int FlowSetProtoFreeFunc (uint8_t , void (*Free)(void *));
void FlowUpdateQueue(Flow *);
int FlowUpdateSpareFlows(void);
static inline void FlowSetNoPacketInspectionFlag(Flow *);
static inline void FlowSetNoPayloadInspectionFlag(Flow *);
int FlowGetPacketDirection(const Flow *, const Packet *);
void FlowCleanupAppLayer(Flow *);
flow-manager: optimize hash walking Until now the flow manager would walk the entire flow hash table on an interval. It would thus touch all flows, leading to a lot of memory and cache pressure. In scenario's where the number of tracked flows run into the hundreds on thousands, and the memory used can run into many hundreds of megabytes or even gigabytes, this would lead to serious performance degradation. This patch introduces a new approach. A timestamp per flow bucket (hash row) is maintained by the flow manager. It holds the timestamp of the earliest possible timeout of a flow in the list. The hash walk skips rows with timestamps beyond the current time. As the timestamp depends on the flows in the hash row's list, and on the 'state' of each flow in the list, any addition of a flow or changing of a flow's state invalidates the timestamp. The flow manager then has to walk the list again to set a new timestamp. A utility function FlowUpdateState is introduced to change Flow states, taking care of the bucket timestamp invalidation while at it. Empty flow buckets use a special value so that we don't have to take the flow bucket lock to find out the bucket is empty. This patch also adds more performance counters: flow_mgr.flows_checked | Total | 929 flow_mgr.flows_notimeout | Total | 391 flow_mgr.flows_timeout | Total | 538 flow_mgr.flows_removed | Total | 277 flow_mgr.flows_timeout_inuse | Total | 261 flow_mgr.rows_checked | Total | 1000000 flow_mgr.rows_skipped | Total | 998835 flow_mgr.rows_empty | Total | 290 flow_mgr.rows_maxlen | Total | 2 flow_mgr.flows_checked: number of flows checked for timeout in the last pass flow_mgr.flows_notimeout: number of flows out of flow_mgr.flows_checked that didn't time out flow_mgr.flows_timeout: number of out of flow_mgr.flows_checked that did reach the time out flow_mgr.flows_removed: number of flows out of flow_mgr.flows_timeout that were really removed flow_mgr.flows_timeout_inuse: number of flows out of flow_mgr.flows_timeout that were still in use or needed work flow_mgr.rows_checked: hash table rows checked flow_mgr.rows_skipped: hash table rows skipped because non of the flows would time out anyway The counters below are only relating to rows that were not skipped. flow_mgr.rows_empty: empty hash rows flow_mgr.rows_maxlen: max number of flows per hash row. Best to keep low, so increase hash-size if needed. flow_mgr.rows_busy: row skipped because it was locked by another thread
9 years ago
void FlowUpdateState(Flow *f, enum FlowState s);
int FlowSetMemcap(uint64_t size);
uint64_t FlowGetMemcap(void);
uint64_t FlowGetMemuse(void);
FlowStorageId GetFlowBypassInfoID(void);
void RegisterFlowBypassInfo(void);
void FlowGetLastTimeAsParts(Flow *flow, uint64_t *secs, uint64_t *usecs);
uint32_t FlowGetFlags(Flow *flow);
uint16_t FlowGetSourcePort(Flow *flow);
uint16_t FlowGetDestinationPort(Flow *flow);
flow-manager: optimize hash walking Until now the flow manager would walk the entire flow hash table on an interval. It would thus touch all flows, leading to a lot of memory and cache pressure. In scenario's where the number of tracked flows run into the hundreds on thousands, and the memory used can run into many hundreds of megabytes or even gigabytes, this would lead to serious performance degradation. This patch introduces a new approach. A timestamp per flow bucket (hash row) is maintained by the flow manager. It holds the timestamp of the earliest possible timeout of a flow in the list. The hash walk skips rows with timestamps beyond the current time. As the timestamp depends on the flows in the hash row's list, and on the 'state' of each flow in the list, any addition of a flow or changing of a flow's state invalidates the timestamp. The flow manager then has to walk the list again to set a new timestamp. A utility function FlowUpdateState is introduced to change Flow states, taking care of the bucket timestamp invalidation while at it. Empty flow buckets use a special value so that we don't have to take the flow bucket lock to find out the bucket is empty. This patch also adds more performance counters: flow_mgr.flows_checked | Total | 929 flow_mgr.flows_notimeout | Total | 391 flow_mgr.flows_timeout | Total | 538 flow_mgr.flows_removed | Total | 277 flow_mgr.flows_timeout_inuse | Total | 261 flow_mgr.rows_checked | Total | 1000000 flow_mgr.rows_skipped | Total | 998835 flow_mgr.rows_empty | Total | 290 flow_mgr.rows_maxlen | Total | 2 flow_mgr.flows_checked: number of flows checked for timeout in the last pass flow_mgr.flows_notimeout: number of flows out of flow_mgr.flows_checked that didn't time out flow_mgr.flows_timeout: number of out of flow_mgr.flows_checked that did reach the time out flow_mgr.flows_removed: number of flows out of flow_mgr.flows_timeout that were really removed flow_mgr.flows_timeout_inuse: number of flows out of flow_mgr.flows_timeout that were still in use or needed work flow_mgr.rows_checked: hash table rows checked flow_mgr.rows_skipped: hash table rows skipped because non of the flows would time out anyway The counters below are only relating to rows that were not skipped. flow_mgr.rows_empty: empty hash rows flow_mgr.rows_maxlen: max number of flows per hash row. Best to keep low, so increase hash-size if needed. flow_mgr.rows_busy: row skipped because it was locked by another thread
9 years ago
/** ----- Inline functions ----- */
/** \brief Set the No Packet Inspection Flag without locking the flow.
*
* \param f Flow to set the flag in
*/
static inline void FlowSetNoPacketInspectionFlag(Flow *f)
{
SCEnter();
SCLogDebug("flow %p", f);
f->flags |= FLOW_NOPACKET_INSPECTION;
SCReturn;
}
/** \brief Set the No payload inspection Flag without locking the flow.
*
* \param f Flow to set the flag in
*/
static inline void FlowSetNoPayloadInspectionFlag(Flow *f)
{
SCEnter();
SCLogDebug("flow %p", f);
f->flags |= FLOW_NOPAYLOAD_INSPECTION;
SCReturn;
}
/**
* \brief increase the use count of a flow
*
* \param f flow to decrease use count for
*/
static inline void FlowIncrUsecnt(Flow *f)
{
if (f == NULL)
return;
flow: redesign of flow timeout handling Goals: - reduce locking - take advantage of 'hot' caches - better locality Locking reduction New flow spare pool. The global pool is implmented as a list of blocks, where each block has a 100 spare flows. Worker threads fetch a block at a time, storing the block in the local thread storage. Flow Recycler now returns flows to the pool is blocks as well. Flow Recycler fetches all flows to be processed in one step instead of one at a time. Cache 'hot'ness Worker threads now check the timeout of flows they evaluate during lookup. The worker will have to read the flow into cache anyway, so the added overhead of checking the timeout value is minimal. When a flow is considered timed out, one of 2 things happens: - if the flow is 'owned' by the thread it is handled locally. Handling means checking if the flow needs 'timeout' work. - otherwise, the flow is added to a special 'evicted' list in the flow bucket where it will be picked up by the flow manager. Flow Manager timing By default the flow manager now tries to do passes of the flow hash in smaller steps, where the goal is to do full pass in 8 x the lowest timeout value it has to enforce. So if the lowest timeout value is 30s, a full pass will take 4 minutes. The goal here is to reduce locking overhead and not get in the way of the workers. In emergency mode each pass is full, and lower timeouts are used. Timing of the flow manager is also no longer relying on pthread condition variables, as these generally cause waking up much quicker than the desired timout. Instead a simple (u)sleep loop is used. Both changes reduce the number of hash passes a lot. Emergency behavior In emergency mode there a number of changes to the workers. In this scenario the flow memcap is fully used up and it is unavoidable that some flows won't be tracked. 1. flow spare pool fetches are reduced to once a second. This avoids locking overhead, while the chance of success was very low. 2. getting an active flow directly from the hash skips flows that had very recent activity to avoid the scenario where all flows get only into the NEW state before getting reused. Rather allow some to have a chance of completing. 3. TCP packets that are not SYN packets will not get a used flow, unless stream.midstream is enabled. The goal here is again to avoid evicting active flows unnecessarily. Better Localily Flow Manager injects flows into the worker threads now, instead of one or two packets. Advantage of this is that the worker threads can get packets from their local packet pools, avoiding constant overhead of packets returning to 'foreign' pools. Counters A lot of flow counters have been added and some have been renamed. Overall the worker threads increment 'flow.wrk.*' counters, while the flow manager increments 'flow.mgr.*'. Additionally, none of the counters are snapshots anymore, they all increment over time. The flow.memuse and flow.spare counters are exceptions. Misc FlowQueue has been split into a FlowQueuePrivate (unlocked) and FlowQueue. Flow no longer has 'prev' pointers and used a unified 'next' pointer for both hash and queue use.
6 years ago
f->use_cnt++;
}
/**
* \brief decrease the use count of a flow
*
* \param f flow to decrease use count for
*/
static inline void FlowDecrUsecnt(Flow *f)
{
if (f == NULL)
return;
flow: redesign of flow timeout handling Goals: - reduce locking - take advantage of 'hot' caches - better locality Locking reduction New flow spare pool. The global pool is implmented as a list of blocks, where each block has a 100 spare flows. Worker threads fetch a block at a time, storing the block in the local thread storage. Flow Recycler now returns flows to the pool is blocks as well. Flow Recycler fetches all flows to be processed in one step instead of one at a time. Cache 'hot'ness Worker threads now check the timeout of flows they evaluate during lookup. The worker will have to read the flow into cache anyway, so the added overhead of checking the timeout value is minimal. When a flow is considered timed out, one of 2 things happens: - if the flow is 'owned' by the thread it is handled locally. Handling means checking if the flow needs 'timeout' work. - otherwise, the flow is added to a special 'evicted' list in the flow bucket where it will be picked up by the flow manager. Flow Manager timing By default the flow manager now tries to do passes of the flow hash in smaller steps, where the goal is to do full pass in 8 x the lowest timeout value it has to enforce. So if the lowest timeout value is 30s, a full pass will take 4 minutes. The goal here is to reduce locking overhead and not get in the way of the workers. In emergency mode each pass is full, and lower timeouts are used. Timing of the flow manager is also no longer relying on pthread condition variables, as these generally cause waking up much quicker than the desired timout. Instead a simple (u)sleep loop is used. Both changes reduce the number of hash passes a lot. Emergency behavior In emergency mode there a number of changes to the workers. In this scenario the flow memcap is fully used up and it is unavoidable that some flows won't be tracked. 1. flow spare pool fetches are reduced to once a second. This avoids locking overhead, while the chance of success was very low. 2. getting an active flow directly from the hash skips flows that had very recent activity to avoid the scenario where all flows get only into the NEW state before getting reused. Rather allow some to have a chance of completing. 3. TCP packets that are not SYN packets will not get a used flow, unless stream.midstream is enabled. The goal here is again to avoid evicting active flows unnecessarily. Better Localily Flow Manager injects flows into the worker threads now, instead of one or two packets. Advantage of this is that the worker threads can get packets from their local packet pools, avoiding constant overhead of packets returning to 'foreign' pools. Counters A lot of flow counters have been added and some have been renamed. Overall the worker threads increment 'flow.wrk.*' counters, while the flow manager increments 'flow.mgr.*'. Additionally, none of the counters are snapshots anymore, they all increment over time. The flow.memuse and flow.spare counters are exceptions. Misc FlowQueue has been split into a FlowQueuePrivate (unlocked) and FlowQueue. Flow no longer has 'prev' pointers and used a unified 'next' pointer for both hash and queue use.
6 years ago
f->use_cnt--;
}
/** \brief Reference the flow, bumping the flows use_cnt
* \note This should only be called once for a destination
* pointer */
static inline void FlowReference(Flow **d, Flow *f)
{
if (likely(f != NULL)) {
#ifdef DEBUG_VALIDATION
BUG_ON(*d == f);
#else
if (*d == f)
return;
#endif
FlowIncrUsecnt(f);
*d = f;
}
}
static inline void FlowDeReference(Flow **d)
{
if (likely(*d != NULL)) {
FlowDecrUsecnt(*d);
*d = NULL;
}
}
/** \brief create a flow id that is as unique as possible
* \retval flow_id signed 64bit id
* \note signed because of the signedness of json_integer_t in
* the json output
*/
static inline int64_t FlowGetId(const Flow *f)
{
int64_t id = (int64_t)f->flow_hash << 31 |
(int64_t)(f->startts.tv_sec & 0x0000FFFF) << 16 |
(int64_t)(f->startts.tv_usec & 0x0000FFFF);
/* reduce to 51 bits as Javascript and even JSON often seem to
* max out there. */
id &= 0x7ffffffffffffLL;
return id;
}
flow: redesign of flow timeout handling Goals: - reduce locking - take advantage of 'hot' caches - better locality Locking reduction New flow spare pool. The global pool is implmented as a list of blocks, where each block has a 100 spare flows. Worker threads fetch a block at a time, storing the block in the local thread storage. Flow Recycler now returns flows to the pool is blocks as well. Flow Recycler fetches all flows to be processed in one step instead of one at a time. Cache 'hot'ness Worker threads now check the timeout of flows they evaluate during lookup. The worker will have to read the flow into cache anyway, so the added overhead of checking the timeout value is minimal. When a flow is considered timed out, one of 2 things happens: - if the flow is 'owned' by the thread it is handled locally. Handling means checking if the flow needs 'timeout' work. - otherwise, the flow is added to a special 'evicted' list in the flow bucket where it will be picked up by the flow manager. Flow Manager timing By default the flow manager now tries to do passes of the flow hash in smaller steps, where the goal is to do full pass in 8 x the lowest timeout value it has to enforce. So if the lowest timeout value is 30s, a full pass will take 4 minutes. The goal here is to reduce locking overhead and not get in the way of the workers. In emergency mode each pass is full, and lower timeouts are used. Timing of the flow manager is also no longer relying on pthread condition variables, as these generally cause waking up much quicker than the desired timout. Instead a simple (u)sleep loop is used. Both changes reduce the number of hash passes a lot. Emergency behavior In emergency mode there a number of changes to the workers. In this scenario the flow memcap is fully used up and it is unavoidable that some flows won't be tracked. 1. flow spare pool fetches are reduced to once a second. This avoids locking overhead, while the chance of success was very low. 2. getting an active flow directly from the hash skips flows that had very recent activity to avoid the scenario where all flows get only into the NEW state before getting reused. Rather allow some to have a chance of completing. 3. TCP packets that are not SYN packets will not get a used flow, unless stream.midstream is enabled. The goal here is again to avoid evicting active flows unnecessarily. Better Localily Flow Manager injects flows into the worker threads now, instead of one or two packets. Advantage of this is that the worker threads can get packets from their local packet pools, avoiding constant overhead of packets returning to 'foreign' pools. Counters A lot of flow counters have been added and some have been renamed. Overall the worker threads increment 'flow.wrk.*' counters, while the flow manager increments 'flow.mgr.*'. Additionally, none of the counters are snapshots anymore, they all increment over time. The flow.memuse and flow.spare counters are exceptions. Misc FlowQueue has been split into a FlowQueuePrivate (unlocked) and FlowQueue. Flow no longer has 'prev' pointers and used a unified 'next' pointer for both hash and queue use.
6 years ago
static inline void FlowSetEndFlags(Flow *f)
{
const int state = f->flow_state;
if (state == FLOW_STATE_NEW)
f->flow_end_flags |= FLOW_END_FLAG_STATE_NEW;
else if (state == FLOW_STATE_ESTABLISHED)
f->flow_end_flags |= FLOW_END_FLAG_STATE_ESTABLISHED;
else if (state == FLOW_STATE_CLOSED)
f->flow_end_flags |= FLOW_END_FLAG_STATE_CLOSED;
else if (state == FLOW_STATE_LOCAL_BYPASSED)
f->flow_end_flags |= FLOW_END_FLAG_STATE_BYPASSED;
#ifdef CAPTURE_OFFLOAD
else if (state == FLOW_STATE_CAPTURE_BYPASSED)
f->flow_end_flags = FLOW_END_FLAG_STATE_BYPASSED;
#endif
}
static inline bool FlowIsBypassed(const Flow *f)
{
if (
#ifdef CAPTURE_OFFLOAD
f->flow_state == FLOW_STATE_CAPTURE_BYPASSED ||
#endif
f->flow_state == FLOW_STATE_LOCAL_BYPASSED) {
return true;
}
return false;
}
int FlowClearMemory(Flow *,uint8_t );
AppProto FlowGetAppProtocol(const Flow *f);
void *FlowGetAppState(const Flow *f);
uint8_t FlowGetDisruptionFlags(const Flow *f, uint8_t flags);
App layer API rewritten. The main files in question are: app-layer.[ch], app-layer-detect-proto.[ch] and app-layer-parser.[ch]. Things addressed in this commit: - Brings out a proper separation between protocol detection phase and the parser phase. - The dns app layer now is registered such that we don't use "dnstcp" and "dnsudp" in the rules. A user who previously wrote a rule like this - "alert dnstcp....." or "alert dnsudp....." would now have to use, alert dns (ipproto:tcp;) or alert udp (app-layer-protocol:dns;) or alert ip (ipproto:udp; app-layer-protocol:dns;) The same rules extend to other another such protocol, dcerpc. - The app layer parser api now takes in the ipproto while registering callbacks. - The app inspection/detection engine also takes an ipproto. - All app layer parser functions now take direction as STREAM_TOSERVER or STREAM_TOCLIENT, as opposed to 0 or 1, which was taken by some of the functions. - FlowInitialize() and FlowRecycle() now resets proto to 0. This is needed by unittests, which would try to clean the flow, and that would call the api, AppLayerParserCleanupParserState(), which would try to clean the app state, but the app layer now needs an ipproto to figure out which api to internally call to clean the state, and if the ipproto is 0, it would return without trying to clean the state. - A lot of unittests are now updated where if they are using a flow and they need to use the app layer, we would set a flow ipproto. - The "app-layer" section in the yaml conf has also been updated as well.
12 years ago
void FlowHandlePacketUpdate(Flow *f, Packet *p, ThreadVars *tv, DecodeThreadVars *dtv);
App layer API rewritten. The main files in question are: app-layer.[ch], app-layer-detect-proto.[ch] and app-layer-parser.[ch]. Things addressed in this commit: - Brings out a proper separation between protocol detection phase and the parser phase. - The dns app layer now is registered such that we don't use "dnstcp" and "dnsudp" in the rules. A user who previously wrote a rule like this - "alert dnstcp....." or "alert dnsudp....." would now have to use, alert dns (ipproto:tcp;) or alert udp (app-layer-protocol:dns;) or alert ip (ipproto:udp; app-layer-protocol:dns;) The same rules extend to other another such protocol, dcerpc. - The app layer parser api now takes in the ipproto while registering callbacks. - The app inspection/detection engine also takes an ipproto. - All app layer parser functions now take direction as STREAM_TOSERVER or STREAM_TOCLIENT, as opposed to 0 or 1, which was taken by some of the functions. - FlowInitialize() and FlowRecycle() now resets proto to 0. This is needed by unittests, which would try to clean the flow, and that would call the api, AppLayerParserCleanupParserState(), which would try to clean the app state, but the app layer now needs an ipproto to figure out which api to internally call to clean the state, and if the ipproto is 0, it would return without trying to clean the state. - A lot of unittests are now updated where if they are using a flow and they need to use the app layer, we would set a flow ipproto. - The "app-layer" section in the yaml conf has also been updated as well.
12 years ago
#endif /* __FLOW_H__ */