1*4882a593Smuzhiyun.. SPDX-License-Identifier: GPL-2.0 2*4882a593Smuzhiyun 3*4882a593Smuzhiyun========================= 4*4882a593SmuzhiyunStream Parser (strparser) 5*4882a593Smuzhiyun========================= 6*4882a593Smuzhiyun 7*4882a593SmuzhiyunIntroduction 8*4882a593Smuzhiyun============ 9*4882a593Smuzhiyun 10*4882a593SmuzhiyunThe stream parser (strparser) is a utility that parses messages of an 11*4882a593Smuzhiyunapplication layer protocol running over a data stream. The stream 12*4882a593Smuzhiyunparser works in conjunction with an upper layer in the kernel to provide 13*4882a593Smuzhiyunkernel support for application layer messages. For instance, Kernel 14*4882a593SmuzhiyunConnection Multiplexor (KCM) uses the Stream Parser to parse messages 15*4882a593Smuzhiyunusing a BPF program. 16*4882a593Smuzhiyun 17*4882a593SmuzhiyunThe strparser works in one of two modes: receive callback or general 18*4882a593Smuzhiyunmode. 19*4882a593Smuzhiyun 20*4882a593SmuzhiyunIn receive callback mode, the strparser is called from the data_ready 21*4882a593Smuzhiyuncallback of a TCP socket. Messages are parsed and delivered as they are 22*4882a593Smuzhiyunreceived on the socket. 23*4882a593Smuzhiyun 24*4882a593SmuzhiyunIn general mode, a sequence of skbs are fed to strparser from an 25*4882a593Smuzhiyunoutside source. Message are parsed and delivered as the sequence is 26*4882a593Smuzhiyunprocessed. This modes allows strparser to be applied to arbitrary 27*4882a593Smuzhiyunstreams of data. 28*4882a593Smuzhiyun 29*4882a593SmuzhiyunInterface 30*4882a593Smuzhiyun========= 31*4882a593Smuzhiyun 32*4882a593SmuzhiyunThe API includes a context structure, a set of callbacks, utility 33*4882a593Smuzhiyunfunctions, and a data_ready function for receive callback mode. The 34*4882a593Smuzhiyuncallbacks include a parse_msg function that is called to perform 35*4882a593Smuzhiyunparsing (e.g. BPF parsing in case of KCM), and a rcv_msg function 36*4882a593Smuzhiyunthat is called when a full message has been completed. 37*4882a593Smuzhiyun 38*4882a593SmuzhiyunFunctions 39*4882a593Smuzhiyun========= 40*4882a593Smuzhiyun 41*4882a593Smuzhiyun :: 42*4882a593Smuzhiyun 43*4882a593Smuzhiyun strp_init(struct strparser *strp, struct sock *sk, 44*4882a593Smuzhiyun const struct strp_callbacks *cb) 45*4882a593Smuzhiyun 46*4882a593Smuzhiyun Called to initialize a stream parser. strp is a struct of type 47*4882a593Smuzhiyun strparser that is allocated by the upper layer. sk is the TCP 48*4882a593Smuzhiyun socket associated with the stream parser for use with receive 49*4882a593Smuzhiyun callback mode; in general mode this is set to NULL. Callbacks 50*4882a593Smuzhiyun are called by the stream parser (the callbacks are listed below). 51*4882a593Smuzhiyun 52*4882a593Smuzhiyun :: 53*4882a593Smuzhiyun 54*4882a593Smuzhiyun void strp_pause(struct strparser *strp) 55*4882a593Smuzhiyun 56*4882a593Smuzhiyun Temporarily pause a stream parser. Message parsing is suspended 57*4882a593Smuzhiyun and no new messages are delivered to the upper layer. 58*4882a593Smuzhiyun 59*4882a593Smuzhiyun :: 60*4882a593Smuzhiyun 61*4882a593Smuzhiyun void strp_unpause(struct strparser *strp) 62*4882a593Smuzhiyun 63*4882a593Smuzhiyun Unpause a paused stream parser. 64*4882a593Smuzhiyun 65*4882a593Smuzhiyun :: 66*4882a593Smuzhiyun 67*4882a593Smuzhiyun void strp_stop(struct strparser *strp); 68*4882a593Smuzhiyun 69*4882a593Smuzhiyun strp_stop is called to completely stop stream parser operations. 70*4882a593Smuzhiyun This is called internally when the stream parser encounters an 71*4882a593Smuzhiyun error, and it is called from the upper layer to stop parsing 72*4882a593Smuzhiyun operations. 73*4882a593Smuzhiyun 74*4882a593Smuzhiyun :: 75*4882a593Smuzhiyun 76*4882a593Smuzhiyun void strp_done(struct strparser *strp); 77*4882a593Smuzhiyun 78*4882a593Smuzhiyun strp_done is called to release any resources held by the stream 79*4882a593Smuzhiyun parser instance. This must be called after the stream processor 80*4882a593Smuzhiyun has been stopped. 81*4882a593Smuzhiyun 82*4882a593Smuzhiyun :: 83*4882a593Smuzhiyun 84*4882a593Smuzhiyun int strp_process(struct strparser *strp, struct sk_buff *orig_skb, 85*4882a593Smuzhiyun unsigned int orig_offset, size_t orig_len, 86*4882a593Smuzhiyun size_t max_msg_size, long timeo) 87*4882a593Smuzhiyun 88*4882a593Smuzhiyun strp_process is called in general mode for a stream parser to 89*4882a593Smuzhiyun parse an sk_buff. The number of bytes processed or a negative 90*4882a593Smuzhiyun error number is returned. Note that strp_process does not 91*4882a593Smuzhiyun consume the sk_buff. max_msg_size is maximum size the stream 92*4882a593Smuzhiyun parser will parse. timeo is timeout for completing a message. 93*4882a593Smuzhiyun 94*4882a593Smuzhiyun :: 95*4882a593Smuzhiyun 96*4882a593Smuzhiyun void strp_data_ready(struct strparser *strp); 97*4882a593Smuzhiyun 98*4882a593Smuzhiyun The upper layer calls strp_tcp_data_ready when data is ready on 99*4882a593Smuzhiyun the lower socket for strparser to process. This should be called 100*4882a593Smuzhiyun from a data_ready callback that is set on the socket. Note that 101*4882a593Smuzhiyun maximum messages size is the limit of the receive socket 102*4882a593Smuzhiyun buffer and message timeout is the receive timeout for the socket. 103*4882a593Smuzhiyun 104*4882a593Smuzhiyun :: 105*4882a593Smuzhiyun 106*4882a593Smuzhiyun void strp_check_rcv(struct strparser *strp); 107*4882a593Smuzhiyun 108*4882a593Smuzhiyun strp_check_rcv is called to check for new messages on the socket. 109*4882a593Smuzhiyun This is normally called at initialization of a stream parser 110*4882a593Smuzhiyun instance or after strp_unpause. 111*4882a593Smuzhiyun 112*4882a593SmuzhiyunCallbacks 113*4882a593Smuzhiyun========= 114*4882a593Smuzhiyun 115*4882a593SmuzhiyunThere are six callbacks: 116*4882a593Smuzhiyun 117*4882a593Smuzhiyun :: 118*4882a593Smuzhiyun 119*4882a593Smuzhiyun int (*parse_msg)(struct strparser *strp, struct sk_buff *skb); 120*4882a593Smuzhiyun 121*4882a593Smuzhiyun parse_msg is called to determine the length of the next message 122*4882a593Smuzhiyun in the stream. The upper layer must implement this function. It 123*4882a593Smuzhiyun should parse the sk_buff as containing the headers for the 124*4882a593Smuzhiyun next application layer message in the stream. 125*4882a593Smuzhiyun 126*4882a593Smuzhiyun The skb->cb in the input skb is a struct strp_msg. Only 127*4882a593Smuzhiyun the offset field is relevant in parse_msg and gives the offset 128*4882a593Smuzhiyun where the message starts in the skb. 129*4882a593Smuzhiyun 130*4882a593Smuzhiyun The return values of this function are: 131*4882a593Smuzhiyun 132*4882a593Smuzhiyun ========= =========================================================== 133*4882a593Smuzhiyun >0 indicates length of successfully parsed message 134*4882a593Smuzhiyun 0 indicates more data must be received to parse the message 135*4882a593Smuzhiyun -ESTRPIPE current message should not be processed by the 136*4882a593Smuzhiyun kernel, return control of the socket to userspace which 137*4882a593Smuzhiyun can proceed to read the messages itself 138*4882a593Smuzhiyun other < 0 Error in parsing, give control back to userspace 139*4882a593Smuzhiyun assuming that synchronization is lost and the stream 140*4882a593Smuzhiyun is unrecoverable (application expected to close TCP socket) 141*4882a593Smuzhiyun ========= =========================================================== 142*4882a593Smuzhiyun 143*4882a593Smuzhiyun In the case that an error is returned (return value is less than 144*4882a593Smuzhiyun zero) and the parser is in receive callback mode, then it will set 145*4882a593Smuzhiyun the error on TCP socket and wake it up. If parse_msg returned 146*4882a593Smuzhiyun -ESTRPIPE and the stream parser had previously read some bytes for 147*4882a593Smuzhiyun the current message, then the error set on the attached socket is 148*4882a593Smuzhiyun ENODATA since the stream is unrecoverable in that case. 149*4882a593Smuzhiyun 150*4882a593Smuzhiyun :: 151*4882a593Smuzhiyun 152*4882a593Smuzhiyun void (*lock)(struct strparser *strp) 153*4882a593Smuzhiyun 154*4882a593Smuzhiyun The lock callback is called to lock the strp structure when 155*4882a593Smuzhiyun the strparser is performing an asynchronous operation (such as 156*4882a593Smuzhiyun processing a timeout). In receive callback mode the default 157*4882a593Smuzhiyun function is to lock_sock for the associated socket. In general 158*4882a593Smuzhiyun mode the callback must be set appropriately. 159*4882a593Smuzhiyun 160*4882a593Smuzhiyun :: 161*4882a593Smuzhiyun 162*4882a593Smuzhiyun void (*unlock)(struct strparser *strp) 163*4882a593Smuzhiyun 164*4882a593Smuzhiyun The unlock callback is called to release the lock obtained 165*4882a593Smuzhiyun by the lock callback. In receive callback mode the default 166*4882a593Smuzhiyun function is release_sock for the associated socket. In general 167*4882a593Smuzhiyun mode the callback must be set appropriately. 168*4882a593Smuzhiyun 169*4882a593Smuzhiyun :: 170*4882a593Smuzhiyun 171*4882a593Smuzhiyun void (*rcv_msg)(struct strparser *strp, struct sk_buff *skb); 172*4882a593Smuzhiyun 173*4882a593Smuzhiyun rcv_msg is called when a full message has been received and 174*4882a593Smuzhiyun is queued. The callee must consume the sk_buff; it can 175*4882a593Smuzhiyun call strp_pause to prevent any further messages from being 176*4882a593Smuzhiyun received in rcv_msg (see strp_pause above). This callback 177*4882a593Smuzhiyun must be set. 178*4882a593Smuzhiyun 179*4882a593Smuzhiyun The skb->cb in the input skb is a struct strp_msg. This 180*4882a593Smuzhiyun struct contains two fields: offset and full_len. Offset is 181*4882a593Smuzhiyun where the message starts in the skb, and full_len is the 182*4882a593Smuzhiyun the length of the message. skb->len - offset may be greater 183*4882a593Smuzhiyun then full_len since strparser does not trim the skb. 184*4882a593Smuzhiyun 185*4882a593Smuzhiyun :: 186*4882a593Smuzhiyun 187*4882a593Smuzhiyun int (*read_sock_done)(struct strparser *strp, int err); 188*4882a593Smuzhiyun 189*4882a593Smuzhiyun read_sock_done is called when the stream parser is done reading 190*4882a593Smuzhiyun the TCP socket in receive callback mode. The stream parser may 191*4882a593Smuzhiyun read multiple messages in a loop and this function allows cleanup 192*4882a593Smuzhiyun to occur when exiting the loop. If the callback is not set (NULL 193*4882a593Smuzhiyun in strp_init) a default function is used. 194*4882a593Smuzhiyun 195*4882a593Smuzhiyun :: 196*4882a593Smuzhiyun 197*4882a593Smuzhiyun void (*abort_parser)(struct strparser *strp, int err); 198*4882a593Smuzhiyun 199*4882a593Smuzhiyun This function is called when stream parser encounters an error 200*4882a593Smuzhiyun in parsing. The default function stops the stream parser and 201*4882a593Smuzhiyun sets the error in the socket if the parser is in receive callback 202*4882a593Smuzhiyun mode. The default function can be changed by setting the callback 203*4882a593Smuzhiyun to non-NULL in strp_init. 204*4882a593Smuzhiyun 205*4882a593SmuzhiyunStatistics 206*4882a593Smuzhiyun========== 207*4882a593Smuzhiyun 208*4882a593SmuzhiyunVarious counters are kept for each stream parser instance. These are in 209*4882a593Smuzhiyunthe strp_stats structure. strp_aggr_stats is a convenience structure for 210*4882a593Smuzhiyunaccumulating statistics for multiple stream parser instances. 211*4882a593Smuzhiyunsave_strp_stats and aggregate_strp_stats are helper functions to save 212*4882a593Smuzhiyunand aggregate statistics. 213*4882a593Smuzhiyun 214*4882a593SmuzhiyunMessage assembly limits 215*4882a593Smuzhiyun======================= 216*4882a593Smuzhiyun 217*4882a593SmuzhiyunThe stream parser provide mechanisms to limit the resources consumed by 218*4882a593Smuzhiyunmessage assembly. 219*4882a593Smuzhiyun 220*4882a593SmuzhiyunA timer is set when assembly starts for a new message. In receive 221*4882a593Smuzhiyuncallback mode the message timeout is taken from rcvtime for the 222*4882a593Smuzhiyunassociated TCP socket. In general mode, the timeout is passed as an 223*4882a593Smuzhiyunargument in strp_process. If the timer fires before assembly completes 224*4882a593Smuzhiyunthe stream parser is aborted and the ETIMEDOUT error is set on the TCP 225*4882a593Smuzhiyunsocket if in receive callback mode. 226*4882a593Smuzhiyun 227*4882a593SmuzhiyunIn receive callback mode, message length is limited to the receive 228*4882a593Smuzhiyunbuffer size of the associated TCP socket. If the length returned by 229*4882a593Smuzhiyunparse_msg is greater than the socket buffer size then the stream parser 230*4882a593Smuzhiyunis aborted with EMSGSIZE error set on the TCP socket. Note that this 231*4882a593Smuzhiyunmakes the maximum size of receive skbuffs for a socket with a stream 232*4882a593Smuzhiyunparser to be 2*sk_rcvbuf of the TCP socket. 233*4882a593Smuzhiyun 234*4882a593SmuzhiyunIn general mode the message length limit is passed in as an argument 235*4882a593Smuzhiyunto strp_process. 236*4882a593Smuzhiyun 237*4882a593SmuzhiyunAuthor 238*4882a593Smuzhiyun====== 239*4882a593Smuzhiyun 240*4882a593SmuzhiyunTom Herbert (tom@quantonium.net) 241