[PATCH 0/6] MR10954: nsiproxy.sys: Reuse icmp sockets for multiple requests and support receiving error packets with Linux ping sockets.
From: Paul Gofman <pgofman@codeweavers.com> --- dlls/nsiproxy.sys/icmp_echo.c | 188 +++++++++++++++++++--------------- 1 file changed, 106 insertions(+), 82 deletions(-) diff --git a/dlls/nsiproxy.sys/icmp_echo.c b/dlls/nsiproxy.sys/icmp_echo.c index af1139e2d48..8d7502b2d85 100644 --- a/dlls/nsiproxy.sys/icmp_echo.c +++ b/dlls/nsiproxy.sys/icmp_echo.c @@ -126,20 +126,28 @@ struct icmp_reply_ctx }; struct family_ops; -struct icmp_data + +struct icmp_socket { - LARGE_INTEGER send_time; + LONG ref; int socket; - int cancel_pipe[2]; - unsigned short id; - unsigned short seq; const struct family_ops *ops; struct sockaddr_storage src_storage; - struct sockaddr_storage dst_storage; int src_len; - int dst_len; int hop_limit; + BYTE ttl, tos; BOOL ping_socket; + unsigned short id; +}; + +struct icmp_data +{ + LARGE_INTEGER send_time; + int cancel_pipe[2]; + unsigned short seq; + struct sockaddr_storage dst_storage; + int dst_len; + struct icmp_socket *s; }; #define MAX_HANDLES 256 /* Max number of simultaneous pings - could become dynamic if need be */ @@ -207,7 +215,7 @@ static void ipv4_init_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp_hd icmp_hdr->type = ICMP4_ECHO_REQUEST; icmp_hdr->code = 0; icmp_hdr->checksum = 0; - icmp_hdr->un.echo.id = data->id = getpid() & 0xffff; /* will be overwritten for linux ping socks */ + icmp_hdr->un.echo.id = data->s->id = getpid() & 0xffff; /* will be overwritten for linux ping socks */ icmp_hdr->un.echo.sequence = data->seq = InterlockedIncrement( &icmp_sequence ) & 0xffff; } @@ -217,7 +225,7 @@ static unsigned short chksum( struct icmp_data *icmp_data, BYTE *data, unsigned unsigned int sum = 0, carry = 0; unsigned short check, s; - if (icmp_data->ping_socket) return 0; + if (icmp_data->s->ping_socket) return 0; while (count > 1) { @@ -259,25 +267,25 @@ static int ipv4_set_reply_ip_status( IP_STATUS ip_status, unsigned int bits, voi } } -static void ipv4_set_socket_opts( struct icmp_data *data, struct icmp_send_echo_params *params ) +static void ipv4_set_socket_opts( struct icmp_socket *s ) { int val; - val = params->ttl; - if (val) setsockopt( data->socket, IPPROTO_IP, IP_TTL, &val, sizeof(val) ); - val = params->tos; - if (val) setsockopt( data->socket, IPPROTO_IP, IP_TOS, &val, sizeof(val) ); + val = s->ttl; + if (val) setsockopt( s->socket, IPPROTO_IP, IP_TTL, &val, sizeof(val) ); + val = s->tos; + if (val) setsockopt( s->socket, IPPROTO_IP, IP_TOS, &val, sizeof(val) ); } #ifdef __linux__ -static void ipv4_linux_ping_set_socket_opts( struct icmp_data *data, struct icmp_send_echo_params *params ) +static void ipv4_linux_ping_set_socket_opts( struct icmp_socket *s ) { static const int val = 1; - ipv4_set_socket_opts( data, params ); + ipv4_set_socket_opts( s ); - setsockopt( data->socket, IPPROTO_IP, IP_RECVTTL, &val, sizeof(val) ); - setsockopt( data->socket, IPPROTO_IP, IP_RECVTOS, &val, sizeof(val) ); + setsockopt( s->socket, IPPROTO_IP, IP_RECVTTL, &val, sizeof(val) ); + setsockopt( s->socket, IPPROTO_IP, IP_RECVTOS, &val, sizeof(val) ); } #endif @@ -374,7 +382,7 @@ static int ipv4_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, i switch (icmp->type) { case ICMP4_ECHO_REPLY: - if ((!data->ping_socket && icmp->un.echo.id != data->id) || + if ((!data->s->ping_socket && icmp->un.echo.id != data->s->id) || icmp->un.echo.sequence != data->seq) return -1; ctx->status = IP_SUCCESS; @@ -406,7 +414,7 @@ static int ipv4_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, i return -1; } - if (data->ping_socket) return 0; + if (data->s->ping_socket) return 0; /* Check that the appended packet is really ours - * all handled icmp replies have an 8-byte header @@ -419,7 +427,7 @@ static int ipv4_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, i orig_icmp_hdr = (const struct icmp_hdr *)((const BYTE *)orig_ip_hdr + orig_ip_hdr_len); if (orig_icmp_hdr->type != ICMP4_ECHO_REQUEST || orig_icmp_hdr->code != 0 || - (!data->ping_socket && orig_icmp_hdr->un.echo.id != data->id) || + (!data->s->ping_socket && orig_icmp_hdr->un.echo.id != data->s->id) || orig_icmp_hdr->un.echo.sequence != data->seq) return -1; ctx->status = status; @@ -480,7 +488,7 @@ struct family_ops void (*init_icmp_hdr)( struct icmp_data *data, struct icmp_hdr *icmp_hdr ); unsigned short (*chksum)( struct icmp_data *icmp_data, BYTE *data, unsigned int count ); int (*set_reply_ip_status)( IP_STATUS ip_status, unsigned int bits, void *out ); - void (*set_socket_opts)( struct icmp_data *data, struct icmp_send_echo_params *params ); + void (*set_socket_opts)( struct icmp_socket *s ); int (*reply_buffer_len)( struct icmp_listen_params *params ); BOOL (*parse_ip_hdr)( struct msghdr *msg, int recvd, int *ip_hdr_len, struct icmp_reply_ctx *ctx ); int (*parse_icmp_hdr)( struct icmp_data *data, struct icmp_hdr *icmp, int icmp_len, struct icmp_reply_ctx *ctx ); @@ -523,7 +531,7 @@ static void ipv6_init_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp_hd icmp_hdr->type = ICMP6_ECHO_REQUEST; icmp_hdr->code = 0; icmp_hdr->checksum = 0; - icmp_hdr->un.echo.id = data->id = getpid() & 0xffff; /* will be overwritten for linux ping socks */ + icmp_hdr->un.echo.id = data->s->id; icmp_hdr->un.echo.sequence = data->seq = InterlockedIncrement( &icmp_sequence ) & 0xffff; } @@ -535,7 +543,7 @@ static unsigned short ipv6_chksum( struct icmp_data *icmp_data, BYTE *data, unsi socklen_t slen; int s, ret; - if (icmp_data->ping_socket) return 0; + if (icmp_data->s->ping_socket) return 0; /* Determine source address. Do it on a separate socket or raw socket won't receive ICMP replies * originating not from the destination address. */ @@ -550,7 +558,7 @@ static unsigned short ipv6_chksum( struct icmp_data *icmp_data, BYTE *data, unsi return 0; } } - if (bind( s, (void *)&icmp_data->src_storage, icmp_data->src_len )) + if (bind( s, (void *)&icmp_data->s->src_storage, icmp_data->s->src_len )) { close( s ); return 0; @@ -587,13 +595,13 @@ static int ipv6_set_reply_ip_status( IP_STATUS ip_status, unsigned int bits, voi return sizeof(*reply); } -static void ipv6_set_socket_opts( struct icmp_data *data, struct icmp_send_echo_params *params ) +static void ipv6_set_socket_opts( struct icmp_socket *s ) { #ifdef IPV6_UNICAST_HOPS { - int val = params->hop_limit; + int val = s->hop_limit; - setsockopt( data->socket, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &val, sizeof(val) ); + setsockopt( s->socket, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &val, sizeof(val) ); } #endif } @@ -636,7 +644,7 @@ static int ipv6_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, switch (icmp->type) { case ICMP6_ECHO_REPLY: - if ((!data->ping_socket && icmp->un.echo.id != data->id) || + if ((!data->s->ping_socket && icmp->un.echo.id != data->s->id) || icmp->un.echo.sequence != data->seq) return -1; ctx->status = IP_SUCCESS; @@ -670,7 +678,7 @@ static int ipv6_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, return -1; } - if (data->ping_socket) return 0; + if (data->s->ping_socket) return 0; /* Check that the appended packet is really ours. */ if (icmp_size < sizeof(*icmp) + sizeof(*orig_ip_hdr)) return -1; orig_ip_hdr = (struct ipv6_hdr *)(icmp + 1); @@ -679,7 +687,7 @@ static int ipv6_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, orig_icmp_hdr = (const struct icmp_hdr *)((const BYTE *)orig_ip_hdr + sizeof(*orig_ip_hdr)); if (orig_icmp_hdr->type != ICMP6_ECHO_REQUEST || orig_icmp_hdr->code != 0 || - (!data->ping_socket && orig_icmp_hdr->un.echo.id != data->id) || + (!data->s->ping_socket && orig_icmp_hdr->un.echo.id != data->s->id) || orig_icmp_hdr->un.echo.sequence != data->seq) return -1; ctx->status = status; @@ -782,100 +790,116 @@ static BOOL sockaddr_to_SOCKADDR_INET( const struct sockaddr *in, SOCKADDR_INET return FALSE; } -static NTSTATUS icmp_data_create( ADDRESS_FAMILY win_family, struct icmp_data **icmp_data ) +static void icmp_release_socket( struct icmp_socket *s ) +{ + if (InterlockedDecrement( &s->ref)) return; + if (s->socket >= 0) close( s->socket ); + free( s ); +} + +static void icmp_data_free( struct icmp_data *data ) { + icmp_release_socket( data->s ); + close( data->cancel_pipe[0] ); + close( data->cancel_pipe[1] ); + free( data ); +} + +static NTSTATUS icmp_data_create( struct icmp_send_echo_params *params, struct icmp_data **icmp_data ) +{ + struct sockaddr *src, *dst; struct icmp_data *data; const struct family_ops *ops; - if (win_family == WS_AF_INET6) ops = &ipv6; - else if (win_family == WS_AF_INET) ops = &ipv4; + if (params->dst->si_family == WS_AF_INET6) ops = &ipv6; + else if (params->dst->si_family == WS_AF_INET) ops = &ipv4; else return STATUS_INVALID_PARAMETER; data = malloc( sizeof(*data) ); if (!data) return STATUS_NO_MEMORY; - - data->ping_socket = FALSE; - data->socket = socket( ops->family, SOCK_RAW, ops->icmp_protocol ); - if (data->socket < 0) /* Try a ping-socket */ + if (!(data->s = calloc( 1, sizeof(*data->s) ))) + { + free( data ); + return STATUS_NO_MEMORY; + } + data->s->ref = 1; + data->s->id = getpid() & 0xffff; /* will be overwritten for linux ping socks */ + data->s->ping_socket = FALSE; + data->s->socket = socket( ops->family, SOCK_RAW, ops->icmp_protocol ); + if (data->s->socket < 0) /* Try a ping-socket */ { TRACE( "failed to open raw sock, trying a dgram sock\n" ); - data->socket = socket( ops->family, SOCK_DGRAM, ops->icmp_protocol ); - if (data->socket < 0) + data->s->socket = socket( ops->family, SOCK_DGRAM, ops->icmp_protocol ); + if (data->s->socket < 0) { WARN( "Unable to create socket\n" ); + icmp_release_socket( data->s ); free( data ); return STATUS_ACCESS_DENIED; } #ifdef __linux__ if (ops->family == AF_INET) ops = &ipv4_linux_ping; - data->ping_socket = TRUE; + data->s->ping_socket = TRUE; #endif } if (pipe( data->cancel_pipe )) { - close( data->socket ); + icmp_release_socket( data->s ); free( data ); return STATUS_ACCESS_DENIED; } - data->ops = ops; + data->s->hop_limit = params->hop_limit; + data->s->ttl = params->ttl; + data->s->tos = params->tos; + src = (struct sockaddr *)&data->s->src_storage; + dst = (struct sockaddr *)&data->dst_storage; + data->s->src_len = SOCKADDR_INET_to_sockaddr( params->src, src, sizeof(data->s->src_storage) ); + data->dst_len = SOCKADDR_INET_to_sockaddr( params->dst, dst, sizeof(data->dst_storage) ); + + if (bind( data->s->socket, src, data->s->src_len )) + { + icmp_data_free( data ); + return STATUS_INVALID_ADDRESS_COMPONENT; + } + + data->s->ops = ops; + data->s->ops->set_socket_opts( data->s ); *icmp_data = data; return STATUS_SUCCESS; } -static void icmp_data_free( struct icmp_data *data ) -{ - close( data->socket ); - close( data->cancel_pipe[0] ); - close( data->cancel_pipe[1] ); - free( data ); -} - NTSTATUS icmp_send_echo( void *args ) { struct icmp_send_echo_params *params = args; struct icmp_hdr *icmp_hdr; /* this is the same for both ipv4 and ipv6 */ struct icmp_data *data; int ret; - struct sockaddr *src, *dst; NTSTATUS status; - status = icmp_data_create( params->dst->si_family, &data ); + status = icmp_data_create( params, &data ); if (status) return status; - data->hop_limit = params->hop_limit; - src = (struct sockaddr *)&data->src_storage; - dst = (struct sockaddr *)&data->dst_storage; - data->src_len = SOCKADDR_INET_to_sockaddr( params->src, src, sizeof(data->src_storage) ); - data->dst_len = SOCKADDR_INET_to_sockaddr( params->dst, dst, sizeof(data->dst_storage) ); - - if (bind( data->socket, src, data->src_len )) - { - icmp_data_free( data ); - return STATUS_INVALID_ADDRESS_COMPONENT; - } - - data->ops->set_socket_opts( data, params ); - icmp_hdr = malloc( sizeof(*icmp_hdr) + params->request_size ); if (!icmp_hdr) { icmp_data_free( data ); return STATUS_NO_MEMORY; } - data->ops->init_icmp_hdr( data, icmp_hdr ); + data->s->ops->init_icmp_hdr( data, icmp_hdr ); memcpy( icmp_hdr + 1, params->request, params->request_size ); - icmp_hdr->checksum = data->ops->chksum( data, (BYTE *)icmp_hdr, sizeof(*icmp_hdr) + params->request_size ); + icmp_hdr->checksum = data->s->ops->chksum( data, (BYTE *)icmp_hdr, sizeof(*icmp_hdr) + params->request_size ); NtQueryPerformanceCounter( &data->send_time, NULL ); - ret = sendto( data->socket, icmp_hdr, sizeof(*icmp_hdr) + params->request_size, 0, dst, data->dst_len ); + ret = sendto( data->s->socket, icmp_hdr, sizeof(*icmp_hdr) + params->request_size, 0, + (struct sockaddr *)&data->dst_storage, data->dst_len ); free( icmp_hdr ); if (ret < 0) { TRACE( "sendto() rets %d errno %d\n", ret, errno ); - params->reply_len = data->ops->set_reply_ip_status( errno_to_ip_status( errno ), params->bits, params->reply ); + params->reply_len = data->s->ops->set_reply_ip_status( errno_to_ip_status( errno ), params->bits, params->reply ); icmp_data_free( data ); return STATUS_SUCCESS; } @@ -917,27 +941,27 @@ static NTSTATUS recv_msg( struct icmp_data *data, struct icmp_listen_params *par char *reply_buf; struct icmp_hdr *icmp_hdr; - reply_buf_len = data->ops->reply_buffer_len( params ); + reply_buf_len = data->s->ops->reply_buffer_len( params ); reply_buf = malloc( reply_buf_len ); if (!reply_buf) return STATUS_NO_MEMORY; iov[0].iov_base = reply_buf; iov[0].iov_len = reply_buf_len; - recvd = recvmsg( data->socket, &msg, 0 ); + recvd = recvmsg( data->s->socket, &msg, 0 ); TRACE( "recvmsg() rets %d errno %d addr_len %d iovlen %d msg_flags %x\n", recvd, errno, msg.msg_namelen, (int)iov[0].iov_len, msg.msg_flags ); if (recvd < 0) goto skip; - if (!data->ops->parse_ip_hdr( &msg, recvd, &ip_hdr_len, &ctx )) goto skip; + if (!data->s->ops->parse_ip_hdr( &msg, recvd, &ip_hdr_len, &ctx )) goto skip; if (recvd < ip_hdr_len + sizeof(*icmp_hdr)) goto skip; icmp_hdr = (struct icmp_hdr *)(reply_buf + ip_hdr_len); - if ((ctx.data_size = data->ops->parse_icmp_hdr( data, icmp_hdr, recvd - ip_hdr_len, &ctx )) < 0) goto skip; + if ((ctx.data_size = data->s->ops->parse_icmp_hdr( data, icmp_hdr, recvd - ip_hdr_len, &ctx )) < 0) goto skip; if (ctx.data_size && msg.msg_flags & MSG_TRUNC) { free( reply_buf ); - params->reply_len = data->ops->set_reply_ip_status( IP_GENERAL_FAILURE, params->bits, params->reply ); + params->reply_len = data->s->ops->set_reply_ip_status( IP_GENERAL_FAILURE, params->bits, params->reply ); return STATUS_SUCCESS; } @@ -945,7 +969,7 @@ static NTSTATUS recv_msg( struct icmp_data *data, struct icmp_listen_params *par ctx.round_trip_time = get_rtt( data->send_time ); ctx.data = icmp_hdr + 1; - data->ops->fill_reply( params, &ctx ); + data->s->ops->fill_reply( params, &ctx ); free( reply_buf ); return STATUS_SUCCESS; @@ -966,14 +990,14 @@ NTSTATUS icmp_listen( void *args ) data = handle_data( params->handle ); if (!data) return STATUS_INVALID_PARAMETER; - if (data->dst_storage.ss_family == AF_INET6 && !data->hop_limit) + if (data->dst_storage.ss_family == AF_INET6 && !data->s->hop_limit) { TRACE( "Invalid hop_limit.\n" ); - params->reply_len = data->ops->set_reply_ip_status( IP_GENERAL_FAILURE, params->bits, params->reply ); + params->reply_len = data->s->ops->set_reply_ip_status( IP_GENERAL_FAILURE, params->bits, params->reply ); return STATUS_SUCCESS; } - fds[0].fd = data->socket; + fds[0].fd = data->s->socket; fds[0].events = POLLIN; fds[1].fd = data->cancel_pipe[0]; fds[1].events = POLLIN; @@ -994,11 +1018,11 @@ NTSTATUS icmp_listen( void *args ) if (!ret) /* timeout */ { TRACE( "timeout\n" ); - params->reply_len = data->ops->set_reply_ip_status( IP_REQ_TIMED_OUT, params->bits, params->reply ); + params->reply_len = data->s->ops->set_reply_ip_status( IP_REQ_TIMED_OUT, params->bits, params->reply ); return STATUS_SUCCESS; } /* ret < 0 */ - params->reply_len = data->ops->set_reply_ip_status( errno_to_ip_status( errno ), params->bits, params->reply ); + params->reply_len = data->s->ops->set_reply_ip_status( errno_to_ip_status( errno ), params->bits, params->reply ); return STATUS_SUCCESS; } -- GitLab https://gitlab.winehq.org/wine/wine/-/merge_requests/10954
From: Paul Gofman <pgofman@codeweavers.com> --- dlls/nsiproxy.sys/device.c | 13 +++++++------ dlls/nsiproxy.sys/icmp_echo.c | 20 ++++++++++---------- dlls/nsiproxy.sys/nsi.c | 2 +- dlls/nsiproxy.sys/nsiproxy_private.h | 2 +- dlls/nsiproxy.sys/unix_private.h | 2 +- 5 files changed, 20 insertions(+), 19 deletions(-) diff --git a/dlls/nsiproxy.sys/device.c b/dlls/nsiproxy.sys/device.c index 4595bf8020c..72b7b380290 100644 --- a/dlls/nsiproxy.sys/device.c +++ b/dlls/nsiproxy.sys/device.c @@ -63,7 +63,7 @@ enum unix_calls { icmp_cancel_listen, icmp_close, - icmp_listen, + icmp_get_reply, icmp_send_echo, nsi_enumerate_all_ex, nsi_get_all_parameters_ex, @@ -225,13 +225,13 @@ static int icmp_echo_reply_struct_len( ULONG family, ULONG bits ) return 0; } -static DWORD WINAPI listen_thread_proc( void *arg ) +static DWORD WINAPI icmp_wait_reply( void *arg ) { IRP *irp = arg; IO_STACK_LOCATION *irpsp = IoGetCurrentIrpStackLocation( irp ); struct nsiproxy_icmp_echo *in = irp->AssociatedIrp.SystemBuffer; struct icmp_close_params close_params; - struct icmp_listen_params params; + struct icmp_get_reply_params params; NTSTATUS status; TRACE( "\n" ); @@ -243,8 +243,8 @@ static DWORD WINAPI listen_thread_proc( void *arg ) params.reply = irp->AssociatedIrp.SystemBuffer; params.reply_len = irpsp->Parameters.DeviceIoControl.OutputBufferLength; - status = nsiproxy_call( icmp_listen, ¶ms ); - TRACE( "icmp_listen rets %08lx\n", status ); + status = nsiproxy_call( icmp_get_reply, ¶ms ); + TRACE( "icmp_get_reply rets %08lx\n", status ); EnterCriticalSection( &icmp_echo_completion_cs ); close_params.handle = irp_set_icmp_handle( irp, 0 ); @@ -263,6 +263,7 @@ static DWORD WINAPI listen_thread_proc( void *arg ) static NTSTATUS handle_send_echo( IRP *irp ) { struct nsiproxy_icmp_echo *in = (struct nsiproxy_icmp_echo *)irp->AssociatedIrp.SystemBuffer; + struct icmp_close_params close_params; struct icmp_send_echo_params params; icmp_handle handle; NTSTATUS status; @@ -297,7 +298,7 @@ static NTSTATUS handle_send_echo( IRP *irp ) } IoMarkIrpPending( irp ); irp_set_icmp_handle( irp, handle ); - RtlQueueWorkItem( listen_thread_proc, irp, WT_EXECUTELONGFUNCTION ); + RtlQueueWorkItem( icmp_wait_reply, irp, WT_EXECUTELONGFUNCTION ); return STATUS_PENDING; } diff --git a/dlls/nsiproxy.sys/icmp_echo.c b/dlls/nsiproxy.sys/icmp_echo.c index 8d7502b2d85..55eae0779c2 100644 --- a/dlls/nsiproxy.sys/icmp_echo.c +++ b/dlls/nsiproxy.sys/icmp_echo.c @@ -289,14 +289,14 @@ static void ipv4_linux_ping_set_socket_opts( struct icmp_socket *s ) } #endif -static int ipv4_reply_buffer_len( struct icmp_listen_params *params ) +static int ipv4_reply_buffer_len( struct icmp_get_reply_params *params ) { int struct_len = (params->bits == 32) ? sizeof(struct icmp_echo_reply_32) : sizeof(struct icmp_echo_reply_64); return sizeof(struct ip_hdr) + sizeof(struct icmp_hdr) + params->reply_len - struct_len; } #ifdef __linux__ -static int ipv4_linux_ping_reply_buffer_len( struct icmp_listen_params *params ) +static int ipv4_linux_ping_reply_buffer_len( struct icmp_get_reply_params *params ) { int struct_len = (params->bits == 32) ? sizeof(struct icmp_echo_reply_32) : sizeof(struct icmp_echo_reply_64); return sizeof(struct icmp_hdr) + params->reply_len - struct_len; @@ -434,7 +434,7 @@ static int ipv4_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, i return 0; } -static void ipv4_fill_reply( struct icmp_listen_params *params, struct icmp_reply_ctx *ctx) +static void ipv4_fill_reply( struct icmp_get_reply_params *params, struct icmp_reply_ctx *ctx) { void *options_data; ULONG data_offset; @@ -489,10 +489,10 @@ struct family_ops unsigned short (*chksum)( struct icmp_data *icmp_data, BYTE *data, unsigned int count ); int (*set_reply_ip_status)( IP_STATUS ip_status, unsigned int bits, void *out ); void (*set_socket_opts)( struct icmp_socket *s ); - int (*reply_buffer_len)( struct icmp_listen_params *params ); + int (*reply_buffer_len)( struct icmp_get_reply_params *params ); BOOL (*parse_ip_hdr)( struct msghdr *msg, int recvd, int *ip_hdr_len, struct icmp_reply_ctx *ctx ); int (*parse_icmp_hdr)( struct icmp_data *data, struct icmp_hdr *icmp, int icmp_len, struct icmp_reply_ctx *ctx ); - void (*fill_reply)( struct icmp_listen_params *params, struct icmp_reply_ctx *ctx ); + void (*fill_reply)( struct icmp_get_reply_params *params, struct icmp_reply_ctx *ctx ); }; static const struct family_ops ipv4 = @@ -606,7 +606,7 @@ static void ipv6_set_socket_opts( struct icmp_socket *s ) #endif } -static int ipv6_reply_buffer_len( struct icmp_listen_params *params ) +static int ipv6_reply_buffer_len( struct icmp_get_reply_params *params ) { return sizeof(struct icmp_hdr) + params->reply_len - sizeof(ICMPV6_ECHO_REPLY); } @@ -695,7 +695,7 @@ static int ipv6_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, return 0; } -static void ipv6_fill_reply( struct icmp_listen_params *params, struct icmp_reply_ctx *ctx) +static void ipv6_fill_reply( struct icmp_get_reply_params *params, struct icmp_reply_ctx *ctx) { ICMPV6_ECHO_REPLY *reply = params->reply; @@ -928,7 +928,7 @@ static ULONG get_rtt( LARGE_INTEGER start ) return (now.QuadPart - start.QuadPart) / 10000; } -static NTSTATUS recv_msg( struct icmp_data *data, struct icmp_listen_params *params ) +static NTSTATUS recv_msg( struct icmp_data *data, struct icmp_get_reply_params *params ) { struct sockaddr_storage addr; struct icmp_reply_ctx ctx; @@ -979,9 +979,9 @@ skip: return STATUS_RETRY; } -NTSTATUS icmp_listen( void *args ) +NTSTATUS icmp_get_reply( void *args ) { - struct icmp_listen_params *params = args; + struct icmp_get_reply_params *params = args; struct icmp_data *data; struct pollfd fds[2]; NTSTATUS status; diff --git a/dlls/nsiproxy.sys/nsi.c b/dlls/nsiproxy.sys/nsi.c index 7fe1e1c9274..41aace2c2c4 100644 --- a/dlls/nsiproxy.sys/nsi.c +++ b/dlls/nsiproxy.sys/nsi.c @@ -345,7 +345,7 @@ const unixlib_entry_t __wine_unix_call_funcs[] = { icmp_cancel_listen, icmp_close, - icmp_listen, + icmp_get_reply, icmp_send_echo, unix_nsi_enumerate_all_ex, unix_nsi_get_all_parameters_ex, diff --git a/dlls/nsiproxy.sys/nsiproxy_private.h b/dlls/nsiproxy.sys/nsiproxy_private.h index ab40fa2057a..a15d594bbdd 100644 --- a/dlls/nsiproxy.sys/nsiproxy_private.h +++ b/dlls/nsiproxy.sys/nsiproxy_private.h @@ -30,7 +30,7 @@ struct icmp_close_params icmp_handle handle; }; -struct icmp_listen_params +struct icmp_get_reply_params { icmp_handle handle; void *reply; diff --git a/dlls/nsiproxy.sys/unix_private.h b/dlls/nsiproxy.sys/unix_private.h index 5487de5111d..8a332f2ba07 100644 --- a/dlls/nsiproxy.sys/unix_private.h +++ b/dlls/nsiproxy.sys/unix_private.h @@ -179,5 +179,5 @@ static inline int ascii_strcasecmp( const char *s1, const char *s2 ) NTSTATUS icmp_cancel_listen( void *args ); NTSTATUS icmp_close( void *args ); -NTSTATUS icmp_listen( void *args ); +NTSTATUS icmp_get_reply( void *args ); NTSTATUS icmp_send_echo( void *args ); -- GitLab https://gitlab.winehq.org/wine/wine/-/merge_requests/10954
From: Paul Gofman <pgofman@codeweavers.com> --- dlls/nsiproxy.sys/device.c | 92 ++++--- dlls/nsiproxy.sys/icmp_echo.c | 359 ++++++++++++++++----------- dlls/nsiproxy.sys/nsi.c | 2 - dlls/nsiproxy.sys/nsiproxy_private.h | 2 +- dlls/nsiproxy.sys/unix_private.h | 2 - 5 files changed, 269 insertions(+), 188 deletions(-) diff --git a/dlls/nsiproxy.sys/device.c b/dlls/nsiproxy.sys/device.c index 72b7b380290..b7180d2cb94 100644 --- a/dlls/nsiproxy.sys/device.c +++ b/dlls/nsiproxy.sys/device.c @@ -19,6 +19,7 @@ */ #include <stdarg.h> +#include <assert.h> #include "ntstatus.h" #include "windef.h" @@ -61,8 +62,6 @@ static NTSTATUS nsiproxy_call( unsigned int code, void *args ) enum unix_calls { - icmp_cancel_listen, - icmp_close, icmp_get_reply, icmp_send_echo, nsi_enumerate_all_ex, @@ -184,35 +183,41 @@ static NTSTATUS nsiproxy_get_parameter( IRP *irp ) return status; } -static inline icmp_handle irp_get_icmp_handle( IRP *irp ) +struct icmp_data { - return PtrToUlong( irp->Tail.Overlay.DriverContext[0] ); + icmp_handle handle; + HANDLE completion_event; + HANDLE wait_object; +}; + +static void free_icmp_data( struct icmp_data *data ) +{ + UnregisterWait( data->wait_object ); + CloseHandle( data->completion_event ); + free( data ); +} + +static struct icmp_data *irp_get_icmp_data( IRP *irp ) +{ + return irp->Tail.Overlay.DriverContext[0]; } -static inline icmp_handle irp_set_icmp_handle( IRP *irp, icmp_handle handle ) +static struct icmp_data *irp_set_icmp_data( IRP *irp, struct icmp_data *data ) { - return PtrToUlong( InterlockedExchangePointer( irp->Tail.Overlay.DriverContext, - ULongToPtr( handle ) ) ); + return InterlockedExchangePointer( irp->Tail.Overlay.DriverContext, data ); } DECLARE_CRITICAL_SECTION( icmp_echo_completion_cs ); static void WINAPI icmp_echo_cancel( DEVICE_OBJECT *device, IRP *irp ) { - struct icmp_cancel_listen_params params; + struct icmp_data *data; TRACE( "device %p, irp %p.\n", device, irp ); IoReleaseCancelSpinLock( irp->CancelIrql ); EnterCriticalSection( &icmp_echo_completion_cs ); - - /* If the handle is not set, either the irp is still - in the request queue, in which case the request thread will - cancel it, or the irp has already finished. If the handle - does exist then notify the listen thread. In all cases the irp - will be completed elsewhere. */ - params.handle = irp_get_icmp_handle( irp ); - if (params.handle) nsiproxy_call( icmp_cancel_listen, ¶ms ); + if ((data = irp_get_icmp_data( irp ))) SetEvent( data->completion_event ); LeaveCriticalSection( &icmp_echo_completion_cs ); } @@ -225,49 +230,56 @@ static int icmp_echo_reply_struct_len( ULONG family, ULONG bits ) return 0; } -static DWORD WINAPI icmp_wait_reply( void *arg ) +static void CALLBACK icmp_wait_complete( void *arg, BOOLEAN timed_out ) { IRP *irp = arg; IO_STACK_LOCATION *irpsp = IoGetCurrentIrpStackLocation( irp ); struct nsiproxy_icmp_echo *in = irp->AssociatedIrp.SystemBuffer; - struct icmp_close_params close_params; struct icmp_get_reply_params params; + struct icmp_data *data; NTSTATUS status; - TRACE( "\n" ); + TRACE( "irp %p, timed_out %d.\n", irp, timed_out ); + EnterCriticalSection( &icmp_echo_completion_cs ); + data = irp_set_icmp_data( irp, NULL ); + assert( data ); params.user_reply_ptr = in->user_reply_ptr; - params.handle = irp_get_icmp_handle( irp ); - params.timeout = in->timeout; + params.handle = data->handle; params.bits = in->bits; params.reply = irp->AssociatedIrp.SystemBuffer; params.reply_len = irpsp->Parameters.DeviceIoControl.OutputBufferLength; status = nsiproxy_call( icmp_get_reply, ¶ms ); - TRACE( "icmp_get_reply rets %08lx\n", status ); - - EnterCriticalSection( &icmp_echo_completion_cs ); - close_params.handle = irp_set_icmp_handle( irp, 0 ); - nsiproxy_call( icmp_close, &close_params ); - irp->IoStatus.Status = status; - if (status == STATUS_SUCCESS) - irp->IoStatus.Information = params.reply_len; + irp->IoStatus.Information = 0; + if (irp->Cancel) + { + TRACE( "irp %p cancelled.\n", irp ); + irp->IoStatus.Status = STATUS_CANCELLED; + } else - irp->IoStatus.Information = 0; + { + irp->IoStatus.Status = status; + if (status == STATUS_SUCCESS) + irp->IoStatus.Information = params.reply_len; + } LeaveCriticalSection( &icmp_echo_completion_cs ); IoCompleteRequest( irp, IO_NO_INCREMENT ); - return 0; + free_icmp_data( data ); } static NTSTATUS handle_send_echo( IRP *irp ) { struct nsiproxy_icmp_echo *in = (struct nsiproxy_icmp_echo *)irp->AssociatedIrp.SystemBuffer; - struct icmp_close_params close_params; struct icmp_send_echo_params params; - icmp_handle handle; + struct icmp_data *data; NTSTATUS status; + data = calloc( 1, sizeof(*data) ); + data->completion_event = CreateEventW( NULL, TRUE, FALSE, NULL ); + irp_set_icmp_data( irp, data ); + TRACE( "\n" ); params.request = in->data + ((in->opt_size + 3) & ~3); params.request_size = in->req_size; @@ -278,7 +290,8 @@ static NTSTATUS handle_send_echo( IRP *irp ) params.tos = in->tos; params.src = &in->src; params.dst = &in->dst; - params.handle = &handle; + params.completion_event = data->completion_event; + params.handle = &data->handle; status = nsiproxy_call( icmp_send_echo, ¶ms ); TRACE( "icmp_send_echo status %#lx\n", status ); @@ -288,17 +301,24 @@ static NTSTATUS handle_send_echo( IRP *irp ) irp->IoStatus.Status = status; if (status == STATUS_SUCCESS) irp->IoStatus.Information = params.reply_len; + free_icmp_data( data ); return status; } IoSetCancelRoutine( irp, icmp_echo_cancel ); if (irp->Cancel && IoSetCancelRoutine( irp, NULL )) { /* IRP was canceled before we set cancel routine */ + icmp_wait_complete( irp, FALSE ); return STATUS_CANCELLED; } IoMarkIrpPending( irp ); - irp_set_icmp_handle( irp, handle ); - RtlQueueWorkItem( icmp_wait_reply, irp, WT_EXECUTELONGFUNCTION ); + if (in->dst.si_family == AF_INET6 && !in->hop_limit) + { + /* This should fail at once but asynchronously. */ + SetEvent( data->completion_event ); + } + RegisterWaitForSingleObject( &data->wait_object, data->completion_event, icmp_wait_complete, irp, in->timeout, + WT_EXECUTEONLYONCE | WT_EXECUTEINWAITTHREAD ); return STATUS_PENDING; } diff --git a/dlls/nsiproxy.sys/icmp_echo.c b/dlls/nsiproxy.sys/icmp_echo.c index 55eae0779c2..6d0473b0105 100644 --- a/dlls/nsiproxy.sys/icmp_echo.c +++ b/dlls/nsiproxy.sys/icmp_echo.c @@ -31,6 +31,7 @@ #include <pthread.h> #include <poll.h> #include <sys/socket.h> +#include <assert.h> #ifdef HAVE_NETINET_IN_H #include <netinet/in.h> @@ -52,6 +53,7 @@ #include "ipexport.h" #include "ipmib.h" #include "wine/nsi.h" +#include "wine/list.h" #include "wine/debug.h" #include "nsiproxy_private.h" @@ -121,14 +123,19 @@ struct icmp_reply_ctx BYTE tos; BYTE flags; BYTE options_size; - void *options_data; - void *data; + unsigned int options_data_offset; + unsigned int data_offset; + BYTE packet[65536]; + unsigned int packet_size; }; struct family_ops; struct icmp_socket { + struct list entry; + struct list listen_entry; + struct list request_list; LONG ref; int socket; const struct family_ops *ops; @@ -138,23 +145,36 @@ struct icmp_socket BYTE ttl, tos; BOOL ping_socket; unsigned short id; + BOOL polling; }; struct icmp_data { + struct list entry; LARGE_INTEGER send_time; - int cancel_pipe[2]; unsigned short seq; struct sockaddr_storage dst_storage; int dst_len; struct icmp_socket *s; + struct icmp_reply_ctx reply_ctx; + HANDLE completion_event; + BOOL polling; + NTSTATUS status; }; + +static pthread_mutex_t listen_lock = PTHREAD_MUTEX_INITIALIZER; +struct list socket_list = LIST_INIT( socket_list ); +static int socket_list_update_pipe[2]; +static pthread_once_t init_once = PTHREAD_ONCE_INIT; + #define MAX_HANDLES 256 /* Max number of simultaneous pings - could become dynamic if need be */ static struct icmp_data *handle_table[MAX_HANDLES]; static pthread_mutex_t handle_lock = PTHREAD_MUTEX_INITIALIZER; static struct icmp_data **next_free, **next_unused = handle_table; +static void icmp_listen( void *args ); + static icmp_handle handle_alloc( struct icmp_data *data ) { struct icmp_data **entry; @@ -289,20 +309,6 @@ static void ipv4_linux_ping_set_socket_opts( struct icmp_socket *s ) } #endif -static int ipv4_reply_buffer_len( struct icmp_get_reply_params *params ) -{ - int struct_len = (params->bits == 32) ? sizeof(struct icmp_echo_reply_32) : sizeof(struct icmp_echo_reply_64); - return sizeof(struct ip_hdr) + sizeof(struct icmp_hdr) + params->reply_len - struct_len; -} - -#ifdef __linux__ -static int ipv4_linux_ping_reply_buffer_len( struct icmp_get_reply_params *params ) -{ - int struct_len = (params->bits == 32) ? sizeof(struct icmp_echo_reply_32) : sizeof(struct icmp_echo_reply_64); - return sizeof(struct icmp_hdr) + params->reply_len - struct_len; -} -#endif - static BOOL ipv4_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip_hdr_len, struct icmp_reply_ctx *ctx ) { @@ -313,7 +319,7 @@ static BOOL ipv4_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip_hdr_len, if (ip_hdr->v_hl >> 4 != 4 || ip_hdr->protocol != IPPROTO_ICMP) return FALSE; *ip_hdr_len = (ip_hdr->v_hl & 0xf) << 2; if (*ip_hdr_len < sizeof(*ip_hdr)) return FALSE; - ctx->options_data = ip_hdr + 1; + ctx->options_data_offset = sizeof(*ip_hdr); ctx->ttl = ip_hdr->ttl; ctx->tos = ip_hdr->tos; ctx->flags = ip_hdr->frag_off >> 13; @@ -329,7 +335,7 @@ static BOOL ipv4_linux_ping_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip struct cmsghdr *cmsg; *ip_hdr_len = 0; - ctx->options_data = NULL; + ctx->options_data_offset = 0; ctx->ttl = 0; ctx->tos = 0; ctx->flags = 0; @@ -434,7 +440,7 @@ static int ipv4_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, i return 0; } -static void ipv4_fill_reply( struct icmp_get_reply_params *params, struct icmp_reply_ctx *ctx) +static BOOL ipv4_fill_reply( struct icmp_get_reply_params *params, struct icmp_reply_ctx *ctx) { void *options_data; ULONG data_offset; @@ -473,12 +479,17 @@ static void ipv4_fill_reply( struct icmp_get_reply_params *params, struct icmp_r options_data = reply + 1; } - memcpy( options_data, ctx->options_data, ctx->options_size ); + if (ctx->options_size && ((char *)options_data - (char *)params->reply) + ctx->options_size > params->reply_len) + return FALSE; + memcpy( options_data, ctx->packet + ctx->options_data_offset, ctx->options_size ); if (ctx->options_size & 3) memset( (char *)options_data + ctx->options_size, 0, 4 - (ctx->options_size & 3) ); - memcpy( (char *)params->reply + data_offset, ctx->data, ctx->data_size ); + if (ctx->data_size && data_offset + ctx->data_size > params->reply_len) + return FALSE; + memcpy( (char *)params->reply + data_offset, ctx->packet + ctx->data_offset, ctx->data_size ); params->reply_len = data_offset + ctx->data_size; + return TRUE; } struct family_ops @@ -489,10 +500,9 @@ struct family_ops unsigned short (*chksum)( struct icmp_data *icmp_data, BYTE *data, unsigned int count ); int (*set_reply_ip_status)( IP_STATUS ip_status, unsigned int bits, void *out ); void (*set_socket_opts)( struct icmp_socket *s ); - int (*reply_buffer_len)( struct icmp_get_reply_params *params ); BOOL (*parse_ip_hdr)( struct msghdr *msg, int recvd, int *ip_hdr_len, struct icmp_reply_ctx *ctx ); int (*parse_icmp_hdr)( struct icmp_data *data, struct icmp_hdr *icmp, int icmp_len, struct icmp_reply_ctx *ctx ); - void (*fill_reply)( struct icmp_get_reply_params *params, struct icmp_reply_ctx *ctx ); + BOOL (*fill_reply)( struct icmp_get_reply_params *params, struct icmp_reply_ctx *ctx ); }; static const struct family_ops ipv4 = @@ -503,7 +513,6 @@ static const struct family_ops ipv4 = chksum, ipv4_set_reply_ip_status, ipv4_set_socket_opts, - ipv4_reply_buffer_len, ipv4_parse_ip_hdr, ipv4_parse_icmp_hdr, ipv4_fill_reply, @@ -519,7 +528,6 @@ static const struct family_ops ipv4_linux_ping = chksum, ipv4_set_reply_ip_status, ipv4_linux_ping_set_socket_opts, - ipv4_linux_ping_reply_buffer_len, ipv4_linux_ping_parse_ip_hdr, ipv4_parse_icmp_hdr, ipv4_fill_reply, @@ -606,16 +614,11 @@ static void ipv6_set_socket_opts( struct icmp_socket *s ) #endif } -static int ipv6_reply_buffer_len( struct icmp_get_reply_params *params ) -{ - return sizeof(struct icmp_hdr) + params->reply_len - sizeof(ICMPV6_ECHO_REPLY); -} - static BOOL ipv6_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip_hdr_len, struct icmp_reply_ctx *ctx ) { *ip_hdr_len = 0; - ctx->options_data = NULL; + ctx->options_data_offset = 0; ctx->ttl = 0; ctx->tos = 0; ctx->flags = 0; @@ -695,7 +698,7 @@ static int ipv6_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, return 0; } -static void ipv6_fill_reply( struct icmp_get_reply_params *params, struct icmp_reply_ctx *ctx) +static BOOL ipv6_fill_reply( struct icmp_get_reply_params *params, struct icmp_reply_ctx *ctx) { ICMPV6_ECHO_REPLY *reply = params->reply; @@ -705,8 +708,11 @@ static void ipv6_fill_reply( struct icmp_get_reply_params *params, struct icmp_r reply->Address.sin6_port = ctx->addr.Ipv6.sin6_port; reply->Address.sin6_scope_id = ctx->addr.Ipv6.sin6_scope_id; reply->RoundTripTime = ctx->round_trip_time; - memcpy( reply + 1, ctx->data, ctx->data_size ); + if (sizeof(*reply) + ctx->data_size > params->reply_len) + return FALSE; + memcpy( reply + 1, ctx->packet + ctx->data_offset, ctx->data_size ); params->reply_len = sizeof(*reply) + ctx->data_size; + return TRUE; } static const struct family_ops ipv6 = @@ -717,7 +723,6 @@ static const struct family_ops ipv6 = ipv6_chksum, ipv6_set_reply_ip_status, ipv6_set_socket_opts, - ipv6_reply_buffer_len, ipv6_parse_ip_hdr, ipv6_parse_icmp_hdr, ipv6_fill_reply, @@ -790,32 +795,72 @@ static BOOL sockaddr_to_SOCKADDR_INET( const struct sockaddr *in, SOCKADDR_INET return FALSE; } +static void icmp_grab_socket( struct icmp_socket *s ) +{ + assert( s->ref > 0 ); + InterlockedIncrement( &s->ref ); +} + static void icmp_release_socket( struct icmp_socket *s ) { - if (InterlockedDecrement( &s->ref)) return; - if (s->socket >= 0) close( s->socket ); - free( s ); + pthread_mutex_lock( &listen_lock ); + if (!InterlockedDecrement( &s->ref )) + { + TRACE( "freeing socket %p, fd %d.\n", s, s->socket ); + if (s->polling) list_remove( &s->entry ); + if (s->socket >= 0) close( s->socket ); + free( s ); + } + pthread_mutex_unlock( &listen_lock ); +} + +static BOOL icmp_data_stop_polling( struct icmp_data *data, NTSTATUS status ) +{ + BOOL ret; + + pthread_mutex_lock( &listen_lock ); + if ((ret = data->polling)) + { + list_remove( &data->entry ); + data->polling = FALSE; + data->status = status; + } + pthread_mutex_unlock( &listen_lock ); + return ret; } static void icmp_data_free( struct icmp_data *data ) { + assert( !data->polling ); icmp_release_socket( data->s ); - close( data->cancel_pipe[0] ); - close( data->cancel_pipe[1] ); free( data ); } +static void init_listen(void) +{ + HANDLE thread; + + pipe( socket_list_update_pipe ); + + if (PsCreateSystemThread( &thread, THREAD_ALL_ACCESS, NULL, 0, NULL, icmp_listen, NULL )) + ERR( "Could not create listen thread.\n" ); + else + NtClose( thread ); +} + static NTSTATUS icmp_data_create( struct icmp_send_echo_params *params, struct icmp_data **icmp_data ) { struct sockaddr *src, *dst; struct icmp_data *data; const struct family_ops *ops; + pthread_once( &init_once, init_listen ); + if (params->dst->si_family == WS_AF_INET6) ops = &ipv6; else if (params->dst->si_family == WS_AF_INET) ops = &ipv4; else return STATUS_INVALID_PARAMETER; - data = malloc( sizeof(*data) ); + data = calloc( 1, sizeof(*data) ); if (!data) return STATUS_NO_MEMORY; if (!(data->s = calloc( 1, sizeof(*data->s) ))) { @@ -823,6 +868,7 @@ static NTSTATUS icmp_data_create( struct icmp_send_echo_params *params, struct i return STATUS_NO_MEMORY; } data->s->ref = 1; + list_init( &data->s->request_list ); data->s->id = getpid() & 0xffff; /* will be overwritten for linux ping socks */ data->s->ping_socket = FALSE; data->s->socket = socket( ops->family, SOCK_RAW, ops->icmp_protocol ); @@ -842,12 +888,6 @@ static NTSTATUS icmp_data_create( struct icmp_send_echo_params *params, struct i data->s->ping_socket = TRUE; #endif } - if (pipe( data->cancel_pipe )) - { - icmp_release_socket( data->s ); - free( data ); - return STATUS_ACCESS_DENIED; - } data->s->hop_limit = params->hop_limit; data->s->ttl = params->ttl; @@ -865,7 +905,14 @@ static NTSTATUS icmp_data_create( struct icmp_send_echo_params *params, struct i data->s->ops = ops; data->s->ops->set_socket_opts( data->s ); + pthread_mutex_lock( &listen_lock ); + list_add_tail( &socket_list, &data->s->entry ); + data->s->polling = TRUE; + write( socket_list_update_pipe[1], "x", 1 ); + pthread_mutex_unlock( &listen_lock ); + *icmp_data = data; + data->completion_event = params->completion_event; return STATUS_SUCCESS; } @@ -874,7 +921,7 @@ NTSTATUS icmp_send_echo( void *args ) struct icmp_send_echo_params *params = args; struct icmp_hdr *icmp_hdr; /* this is the same for both ipv4 and ipv6 */ struct icmp_data *data; - int ret; + int ret, err; NTSTATUS status; @@ -891,35 +938,36 @@ NTSTATUS icmp_send_echo( void *args ) memcpy( icmp_hdr + 1, params->request, params->request_size ); icmp_hdr->checksum = data->s->ops->chksum( data, (BYTE *)icmp_hdr, sizeof(*icmp_hdr) + params->request_size ); + pthread_mutex_lock( &listen_lock ); + list_add_tail( &data->s->request_list, &data->entry ); + data->polling = TRUE; + pthread_mutex_unlock( &listen_lock ); + NtQueryPerformanceCounter( &data->send_time, NULL ); ret = sendto( data->s->socket, icmp_hdr, sizeof(*icmp_hdr) + params->request_size, 0, (struct sockaddr *)&data->dst_storage, data->dst_len ); + if (ret < 0) err = errno; free( icmp_hdr ); if (ret < 0) { - TRACE( "sendto() rets %d errno %d\n", ret, errno ); - params->reply_len = data->s->ops->set_reply_ip_status( errno_to_ip_status( errno ), params->bits, params->reply ); + status = errno_to_ip_status( err ); + TRACE( "sendto() rets %d errno %d\n", ret, err ); + icmp_data_stop_polling( data, status ); + params->reply_len = data->s->ops->set_reply_ip_status( status, params->bits, params->reply ); icmp_data_free( data ); return STATUS_SUCCESS; } *params->handle = handle_alloc( data ); - if (!*params->handle) icmp_data_free( data ); + if (!*params->handle) + { + icmp_data_stop_polling( data, STATUS_NO_MEMORY ); + icmp_data_free( data ); + } return *params->handle ? STATUS_PENDING : STATUS_NO_MEMORY; } -static int get_timeout( LARGE_INTEGER start, UINT timeout ) -{ - LARGE_INTEGER now, end; - - end.QuadPart = start.QuadPart + (ULONGLONG)timeout * 10000; - NtQueryPerformanceCounter( &now, NULL ); - if (now.QuadPart >= end.QuadPart) return 0; - - return min( (end.QuadPart - now.QuadPart) / 10000, INT_MAX ); -} - static ULONG get_rtt( LARGE_INTEGER start ) { LARGE_INTEGER now; @@ -928,121 +976,138 @@ static ULONG get_rtt( LARGE_INTEGER start ) return (now.QuadPart - start.QuadPart) / 10000; } -static NTSTATUS recv_msg( struct icmp_data *data, struct icmp_get_reply_params *params ) +static void icmp_listen( void *args ) { struct sockaddr_storage addr; struct icmp_reply_ctx ctx; - struct iovec iov[1]; BYTE cmsg_buf[1024]; - struct msghdr msg = { .msg_name = &addr, .msg_namelen = sizeof(addr), - .msg_iov = iov, .msg_iovlen = ARRAY_SIZE(iov), - .msg_control = cmsg_buf, .msg_controllen = sizeof(cmsg_buf) }; - int ip_hdr_len, recvd, reply_buf_len; - char *reply_buf; + struct iovec iov[1] = {{ .iov_base = ctx.packet, .iov_len = sizeof(ctx.packet) }}; + struct msghdr msg = { .msg_iov = iov, .msg_iovlen = ARRAY_SIZE(iov) }; + struct list listen_list; + unsigned int fds_size = 64; + struct pollfd *fds = malloc( fds_size * sizeof(*fds) ); + struct icmp_socket *s, *next; struct icmp_hdr *icmp_hdr; + int ip_hdr_len, recvd; + struct icmp_data *data; + unsigned int i, count; + int ret; + BYTE b; - reply_buf_len = data->s->ops->reply_buffer_len( params ); - reply_buf = malloc( reply_buf_len ); - if (!reply_buf) return STATUS_NO_MEMORY; - - iov[0].iov_base = reply_buf; - iov[0].iov_len = reply_buf_len; - - recvd = recvmsg( data->s->socket, &msg, 0 ); - TRACE( "recvmsg() rets %d errno %d addr_len %d iovlen %d msg_flags %x\n", - recvd, errno, msg.msg_namelen, (int)iov[0].iov_len, msg.msg_flags ); - - if (recvd < 0) goto skip; - if (!data->s->ops->parse_ip_hdr( &msg, recvd, &ip_hdr_len, &ctx )) goto skip; - if (recvd < ip_hdr_len + sizeof(*icmp_hdr)) goto skip; - - icmp_hdr = (struct icmp_hdr *)(reply_buf + ip_hdr_len); - if ((ctx.data_size = data->s->ops->parse_icmp_hdr( data, icmp_hdr, recvd - ip_hdr_len, &ctx )) < 0) goto skip; - if (ctx.data_size && msg.msg_flags & MSG_TRUNC) + pthread_once( &init_once, init_listen ); + while (1) { - free( reply_buf ); - params->reply_len = data->s->ops->set_reply_ip_status( IP_GENERAL_FAILURE, params->bits, params->reply ); - return STATUS_SUCCESS; - } + list_init( &listen_list ); - sockaddr_to_SOCKADDR_INET( (struct sockaddr *)&addr, &ctx.addr ); - ctx.round_trip_time = get_rtt( data->send_time ); - ctx.data = icmp_hdr + 1; - - data->s->ops->fill_reply( params, &ctx ); - - free( reply_buf ); - return STATUS_SUCCESS; + pthread_mutex_lock( &listen_lock ); + i = 0; + LIST_FOR_EACH_ENTRY( s, &socket_list, struct icmp_socket, entry ) + { + icmp_grab_socket( s ); + list_add_tail( &listen_list, &s->listen_entry ); + if (i + 1 >= fds_size) + { + fds_size *= 2; + fds = realloc( fds, fds_size * sizeof(*fds) ); + } + fds[i].fd = s->socket; + fds[i].events = POLLIN; + ++i; + } + pthread_mutex_unlock( &listen_lock ); + + fds[i].fd = socket_list_update_pipe[0]; + fds[i].events = POLLIN; + count = i; + while ((ret = poll( fds, count + 1, -1 )) < 0 && (errno == EINTR || errno == EAGAIN)) + ; + if (ret < 0) + { + ERR( "poll(): ret %d, %s.\n", ret, strerror( errno )); + return; + } + i = 0; + LIST_FOR_EACH_ENTRY( s, &listen_list, struct icmp_socket, listen_entry ) + { + if (!fds[i].revents & POLLIN) goto skip; + msg.msg_name = &addr; + msg.msg_namelen = sizeof(addr); + msg.msg_control = cmsg_buf; + msg.msg_controllen = sizeof(cmsg_buf); + recvd = recvmsg( s->socket, &msg, 0 ); + TRACE( "s %p, recvmsg() rets %d errno %d addr_len %d iovlen %d msg_flags %x\n", + s, recvd, errno, msg.msg_namelen, (int)iov[0].iov_len, msg.msg_flags ); + if (recvd < 0) goto skip; + if (!s->ops->parse_ip_hdr( &msg, recvd, &ip_hdr_len, &ctx )) goto skip; + if (recvd < ip_hdr_len + sizeof(*icmp_hdr)) goto skip; + ctx.packet_size = recvd; + icmp_hdr = (struct icmp_hdr *)(ctx.packet + ip_hdr_len); + pthread_mutex_lock( &listen_lock ); + LIST_FOR_EACH_ENTRY( data, &s->request_list, struct icmp_data, entry ) + { + if ((ctx.data_size = s->ops->parse_icmp_hdr( data, icmp_hdr, recvd - ip_hdr_len, &ctx )) < 0) + continue; + data->polling = FALSE; + list_remove( &data->entry ); + sockaddr_to_SOCKADDR_INET( (struct sockaddr *)&addr, &ctx.addr ); + ctx.round_trip_time = get_rtt( data->send_time ); + ctx.data_offset = (BYTE *)(icmp_hdr + 1) - ctx.packet; + data->reply_ctx = ctx; + NtSetEvent( data->completion_event, NULL ); + break; + } + pthread_mutex_unlock( &listen_lock ); skip: - free( reply_buf ); - return STATUS_RETRY; + ++i; + } + if (fds[count].revents & POLLIN) + { + read( socket_list_update_pipe[0], &b, 1 ); + TRACE( "updating socket list.\n" ); + } + + LIST_FOR_EACH_ENTRY_SAFE( s, next, &listen_list, struct icmp_socket, listen_entry ) + { + icmp_release_socket( s ); + } + } } NTSTATUS icmp_get_reply( void *args ) { struct icmp_get_reply_params *params = args; struct icmp_data *data; - struct pollfd fds[2]; - NTSTATUS status; - int ret; + NTSTATUS ret; data = handle_data( params->handle ); if (!data) return STATUS_INVALID_PARAMETER; if (data->dst_storage.ss_family == AF_INET6 && !data->s->hop_limit) { - TRACE( "Invalid hop_limit.\n" ); + TRACE( "data %p, invalid hop_limit.\n", data ); + icmp_data_stop_polling( data, STATUS_SUCCESS ); params->reply_len = data->s->ops->set_reply_ip_status( IP_GENERAL_FAILURE, params->bits, params->reply ); return STATUS_SUCCESS; } - fds[0].fd = data->s->socket; - fds[0].events = POLLIN; - fds[1].fd = data->cancel_pipe[0]; - fds[1].events = POLLIN; - - while ((ret = poll( fds, ARRAY_SIZE(fds), get_timeout( data->send_time, params->timeout ) )) > 0) + if (icmp_data_stop_polling( data, STATUS_TIMEOUT )) { - if (fds[1].revents & POLLIN) - { - TRACE( "cancelled\n" ); - return STATUS_CANCELLED; - } - - status = recv_msg( data, params ); - if (status == STATUS_RETRY) continue; - return status; + /* Was still polling, timeout. */ + TRACE( "data %p, timeout.\n", data ); + params->reply_len = data->s->ops->set_reply_ip_status( IP_REQ_TIMED_OUT, params->bits, params->reply ); + ret = STATUS_SUCCESS; } - - if (!ret) /* timeout */ + else if (!(ret = data->status)) { - TRACE( "timeout\n" ); - params->reply_len = data->s->ops->set_reply_ip_status( IP_REQ_TIMED_OUT, params->bits, params->reply ); - return STATUS_SUCCESS; + if (!data->s->ops->fill_reply( params, &data->reply_ctx )) + { + WARN( "data %p, buffer too small.\n", data ); + params->reply_len = data->s->ops->set_reply_ip_status( IP_GENERAL_FAILURE, params->bits, params->reply ); + } + else TRACE( "data %p, seq %u, got reply %u.\n", data, data->seq, data->reply_ctx.status ); } - /* ret < 0 */ - params->reply_len = data->s->ops->set_reply_ip_status( errno_to_ip_status( errno ), params->bits, params->reply ); - return STATUS_SUCCESS; -} - -NTSTATUS icmp_cancel_listen( void *args ) -{ - struct icmp_cancel_listen_params *params = args; - struct icmp_data *data = handle_data( params->handle ); - - if (!data) return STATUS_INVALID_PARAMETER; - write( data->cancel_pipe[1], "x", 1 ); - return STATUS_SUCCESS; -} - -NTSTATUS icmp_close( void *args ) -{ - struct icmp_close_params *params = args; - struct icmp_data *data = handle_data( params->handle ); - - if (!data) return STATUS_INVALID_PARAMETER; icmp_data_free( data ); handle_free( params->handle ); - return STATUS_SUCCESS; + return ret; } diff --git a/dlls/nsiproxy.sys/nsi.c b/dlls/nsiproxy.sys/nsi.c index 41aace2c2c4..09ced8f814e 100644 --- a/dlls/nsiproxy.sys/nsi.c +++ b/dlls/nsiproxy.sys/nsi.c @@ -343,8 +343,6 @@ static NTSTATUS unix_nsi_get_notification( void *args ) const unixlib_entry_t __wine_unix_call_funcs[] = { - icmp_cancel_listen, - icmp_close, icmp_get_reply, icmp_send_echo, unix_nsi_enumerate_all_ex, diff --git a/dlls/nsiproxy.sys/nsiproxy_private.h b/dlls/nsiproxy.sys/nsiproxy_private.h index a15d594bbdd..e30f1a4527b 100644 --- a/dlls/nsiproxy.sys/nsiproxy_private.h +++ b/dlls/nsiproxy.sys/nsiproxy_private.h @@ -36,7 +36,6 @@ struct icmp_get_reply_params void *reply; ULONGLONG user_reply_ptr; unsigned int bits, reply_len; - int timeout; }; struct icmp_send_echo_params @@ -47,6 +46,7 @@ struct icmp_send_echo_params UINT request_size, reply_len; BYTE bits, ttl, tos; int hop_limit; + HANDLE completion_event; icmp_handle *handle; }; diff --git a/dlls/nsiproxy.sys/unix_private.h b/dlls/nsiproxy.sys/unix_private.h index 8a332f2ba07..170bdba665d 100644 --- a/dlls/nsiproxy.sys/unix_private.h +++ b/dlls/nsiproxy.sys/unix_private.h @@ -177,7 +177,5 @@ static inline int ascii_strcasecmp( const char *s1, const char *s2 ) return ascii_strncasecmp( s1, s2, -1 ); } -NTSTATUS icmp_cancel_listen( void *args ); -NTSTATUS icmp_close( void *args ); NTSTATUS icmp_get_reply( void *args ); NTSTATUS icmp_send_echo( void *args ); -- GitLab https://gitlab.winehq.org/wine/wine/-/merge_requests/10954
From: Paul Gofman <pgofman@codeweavers.com> --- dlls/nsiproxy.sys/icmp_echo.c | 62 +++++++++++++++++++++-------------- 1 file changed, 38 insertions(+), 24 deletions(-) diff --git a/dlls/nsiproxy.sys/icmp_echo.c b/dlls/nsiproxy.sys/icmp_echo.c index 6d0473b0105..8bc89ebae4a 100644 --- a/dlls/nsiproxy.sys/icmp_echo.c +++ b/dlls/nsiproxy.sys/icmp_echo.c @@ -127,6 +127,7 @@ struct icmp_reply_ctx unsigned int data_offset; BYTE packet[65536]; unsigned int packet_size; + uint16_t id, sequence; }; struct family_ops; @@ -358,7 +359,7 @@ static BOOL ipv4_linux_ping_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip } #endif -static int ipv4_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, int icmp_size, +static int ipv4_parse_icmp_hdr( struct icmp_socket *s, struct icmp_hdr *icmp, int icmp_size, struct icmp_reply_ctx *ctx ) { static const IP_STATUS unreach_codes[] = @@ -388,9 +389,8 @@ static int ipv4_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, i switch (icmp->type) { case ICMP4_ECHO_REPLY: - if ((!data->s->ping_socket && icmp->un.echo.id != data->s->id) || - icmp->un.echo.sequence != data->seq) return -1; - + ctx->id = icmp->un.echo.id; + ctx->sequence = icmp->un.echo.sequence; ctx->status = IP_SUCCESS; return icmp_size - sizeof(*icmp); @@ -420,11 +420,13 @@ static int ipv4_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, i return -1; } - if (data->s->ping_socket) return 0; + if (s->ping_socket) + { + /* Ping sockets only return echo reply through normal recvmsg() received data, should not get here. */ + return -1; + } - /* Check that the appended packet is really ours - - * all handled icmp replies have an 8-byte header - * followed by the original ip hdr. */ + /* All handled icmp replies have an 8-byte header followed by the original ip hdr. */ if (icmp_size < sizeof(*icmp) + sizeof(*orig_ip_hdr)) return -1; orig_ip_hdr = (struct ip_hdr *)(icmp + 1); if (orig_ip_hdr->v_hl >> 4 != 4 || orig_ip_hdr->protocol != IPPROTO_ICMP) return -1; @@ -432,10 +434,10 @@ static int ipv4_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, i if (icmp_size < sizeof(*icmp) + orig_ip_hdr_len + sizeof(*orig_icmp_hdr)) return -1; orig_icmp_hdr = (const struct icmp_hdr *)((const BYTE *)orig_ip_hdr + orig_ip_hdr_len); if (orig_icmp_hdr->type != ICMP4_ECHO_REQUEST || - orig_icmp_hdr->code != 0 || - (!data->s->ping_socket && orig_icmp_hdr->un.echo.id != data->s->id) || - orig_icmp_hdr->un.echo.sequence != data->seq) return -1; + orig_icmp_hdr->code != 0) return -1; + ctx->id = orig_icmp_hdr->un.echo.id; + ctx->sequence = orig_icmp_hdr->un.echo.sequence; ctx->status = status; return 0; } @@ -501,7 +503,7 @@ struct family_ops int (*set_reply_ip_status)( IP_STATUS ip_status, unsigned int bits, void *out ); void (*set_socket_opts)( struct icmp_socket *s ); BOOL (*parse_ip_hdr)( struct msghdr *msg, int recvd, int *ip_hdr_len, struct icmp_reply_ctx *ctx ); - int (*parse_icmp_hdr)( struct icmp_data *data, struct icmp_hdr *icmp, int icmp_len, struct icmp_reply_ctx *ctx ); + int (*parse_icmp_hdr)( struct icmp_socket *s, struct icmp_hdr *icmp, int icmp_len, struct icmp_reply_ctx *ctx ); BOOL (*fill_reply)( struct icmp_get_reply_params *params, struct icmp_reply_ctx *ctx ); }; @@ -626,7 +628,7 @@ static BOOL ipv6_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip_hdr_len, return TRUE; } -static int ipv6_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, +static int ipv6_parse_icmp_hdr( struct icmp_socket *s, struct icmp_hdr *icmp, int icmp_size, struct icmp_reply_ctx *ctx ) { static const IP_STATUS unreach_codes[] = @@ -647,9 +649,8 @@ static int ipv6_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, switch (icmp->type) { case ICMP6_ECHO_REPLY: - if ((!data->s->ping_socket && icmp->un.echo.id != data->s->id) || - icmp->un.echo.sequence != data->seq) return -1; - + ctx->id = icmp->un.echo.id; + ctx->sequence = icmp->un.echo.sequence; ctx->status = IP_SUCCESS; return icmp_size - sizeof(*icmp); @@ -681,20 +682,24 @@ static int ipv6_parse_icmp_hdr( struct icmp_data *data, struct icmp_hdr *icmp, return -1; } - if (data->s->ping_socket) return 0; - /* Check that the appended packet is really ours. */ + if (s->ping_socket) + { + /* Ping sockets only return echo reply through normal recvmsg() received data, should not get here. */ + return -1; + } + + /* All handled icmp replies have an 8-byte header followed by the original ip hdr. */ if (icmp_size < sizeof(*icmp) + sizeof(*orig_ip_hdr)) return -1; orig_ip_hdr = (struct ipv6_hdr *)(icmp + 1); if ((orig_ip_hdr->v_prio >> 4) != 6 || orig_ip_hdr->next_hdr != IPPROTO_ICMPV6) return -1; if (icmp_size < sizeof(*icmp) + sizeof(*orig_ip_hdr) + sizeof(*orig_icmp_hdr)) return -1; orig_icmp_hdr = (const struct icmp_hdr *)((const BYTE *)orig_ip_hdr + sizeof(*orig_ip_hdr)); if (orig_icmp_hdr->type != ICMP6_ECHO_REQUEST || - orig_icmp_hdr->code != 0 || - (!data->s->ping_socket && orig_icmp_hdr->un.echo.id != data->s->id) || - orig_icmp_hdr->un.echo.sequence != data->seq) return -1; + orig_icmp_hdr->code != 0) return -1; + ctx->id = orig_icmp_hdr->un.echo.id; + ctx->sequence = orig_icmp_hdr->un.echo.sequence; ctx->status = status; - return 0; } @@ -991,6 +996,7 @@ static void icmp_listen( void *args ) int ip_hdr_len, recvd; struct icmp_data *data; unsigned int i, count; + BOOL matched; int ret; BYTE b; @@ -1043,11 +1049,18 @@ static void icmp_listen( void *args ) if (recvd < ip_hdr_len + sizeof(*icmp_hdr)) goto skip; ctx.packet_size = recvd; icmp_hdr = (struct icmp_hdr *)(ctx.packet + ip_hdr_len); + if ((ctx.data_size = s->ops->parse_icmp_hdr( s, icmp_hdr, recvd - ip_hdr_len, &ctx )) < 0) + { + WARN( "parse_icmp_hdr failed.\n" ); + goto skip; + } pthread_mutex_lock( &listen_lock ); + matched = FALSE; LIST_FOR_EACH_ENTRY( data, &s->request_list, struct icmp_data, entry ) { - if ((ctx.data_size = s->ops->parse_icmp_hdr( data, icmp_hdr, recvd - ip_hdr_len, &ctx )) < 0) - continue; + if ((!s->ping_socket && ctx.id != s->id) || ctx.sequence != data->seq) continue; + TRACE( "id %d, seq %d matched to %p.\n", ctx.id, ctx.sequence, data ); + matched = TRUE; data->polling = FALSE; list_remove( &data->entry ); sockaddr_to_SOCKADDR_INET( (struct sockaddr *)&addr, &ctx.addr ); @@ -1058,6 +1071,7 @@ static void icmp_listen( void *args ) break; } pthread_mutex_unlock( &listen_lock ); + if (!matched) TRACE( "id %d, seq %d not matched.\n", ctx.id, ctx.sequence ); skip: ++i; } -- GitLab https://gitlab.winehq.org/wine/wine/-/merge_requests/10954
From: Paul Gofman <pgofman@codeweavers.com> --- dlls/nsiproxy.sys/icmp_echo.c | 117 ++++++++++++++++++++++------------ 1 file changed, 77 insertions(+), 40 deletions(-) diff --git a/dlls/nsiproxy.sys/icmp_echo.c b/dlls/nsiproxy.sys/icmp_echo.c index 8bc89ebae4a..d92b420715e 100644 --- a/dlls/nsiproxy.sys/icmp_echo.c +++ b/dlls/nsiproxy.sys/icmp_echo.c @@ -751,6 +751,7 @@ static int SOCKADDR_INET_to_sockaddr( const SOCKADDR_INET *in, struct sockaddr * struct sockaddr_in *sa = (struct sockaddr_in *)out; if (len < sizeof(*sa)) return 0; + memset( sa, 0, sizeof(*sa) ); sa->sin_family = AF_INET; sa->sin_port = in->Ipv4.sin_port; sa->sin_addr.s_addr = in->Ipv4.sin_addr.WS_s_addr; @@ -761,6 +762,7 @@ static int SOCKADDR_INET_to_sockaddr( const SOCKADDR_INET *in, struct sockaddr * struct sockaddr_in6 *sa = (struct sockaddr_in6 *)out; if (len < sizeof(*sa)) return 0; + memset( sa, 0, sizeof(*sa) ); sa->sin6_family = AF_INET6; sa->sin6_port = in->Ipv6.sin6_port; sa->sin6_flowinfo = in->Ipv6.sin6_flowinfo; @@ -853,71 +855,94 @@ static void init_listen(void) NtClose( thread ); } -static NTSTATUS icmp_data_create( struct icmp_send_echo_params *params, struct icmp_data **icmp_data ) +static NTSTATUS icmp_get_socket( struct icmp_send_echo_params *params, struct icmp_socket **ret_socket ) { - struct sockaddr *src, *dst; - struct icmp_data *data; - const struct family_ops *ops; + struct family_ops *ops; + struct icmp_socket *s; + struct sockaddr *src; + int src_len; + struct sockaddr_storage src_storage; - pthread_once( &init_once, init_listen ); + src = (struct sockaddr *)&src_storage; + src_len = SOCKADDR_INET_to_sockaddr( params->src, src, sizeof(src_storage) ); if (params->dst->si_family == WS_AF_INET6) ops = &ipv6; else if (params->dst->si_family == WS_AF_INET) ops = &ipv4; else return STATUS_INVALID_PARAMETER; - data = calloc( 1, sizeof(*data) ); - if (!data) return STATUS_NO_MEMORY; - if (!(data->s = calloc( 1, sizeof(*data->s) ))) + *ret_socket = NULL; + pthread_mutex_lock( &listen_lock ); + LIST_FOR_EACH_ENTRY( s, &socket_list, struct icmp_socket, entry ) { - free( data ); + assert( s->polling ); + if (params->dst->si_family == WS_AF_INET6 && ops->family != AF_INET6) continue; + if (params->dst->si_family == WS_AF_INET && ops->family != AF_INET) continue; + if (s->hop_limit != params->hop_limit || s->ttl != params->ttl || s->tos != params->tos) continue; + if (s->src_len != src_len) continue; + if (memcmp( src, &s->src_storage, src_len )) continue; + + icmp_grab_socket( s ); + pthread_mutex_unlock( &listen_lock ); + *ret_socket = s; + TRACE( "Using socket %p.\n", s ); + return STATUS_SUCCESS; + } + + pthread_once( &init_once, init_listen ); + + if (!(s = calloc( 1, sizeof(*s) ))) + { + pthread_mutex_unlock( &listen_lock ); return STATUS_NO_MEMORY; } - data->s->ref = 1; - list_init( &data->s->request_list ); - data->s->id = getpid() & 0xffff; /* will be overwritten for linux ping socks */ - data->s->ping_socket = FALSE; - data->s->socket = socket( ops->family, SOCK_RAW, ops->icmp_protocol ); - if (data->s->socket < 0) /* Try a ping-socket */ + s->ref = 1; + list_init( &s->request_list ); + s->socket = -1; + s->id = getpid() & 0xffff; /* will be overwritten for linux ping socks */ + s->ping_socket = FALSE; + + s->socket = socket( ops->family, SOCK_RAW, ops->icmp_protocol ); + if (s->socket < 0) /* Try a ping-socket */ { + int value = 1; + TRACE( "failed to open raw sock, trying a dgram sock\n" ); - data->s->socket = socket( ops->family, SOCK_DGRAM, ops->icmp_protocol ); - if (data->s->socket < 0) + s->socket = socket( ops->family, SOCK_DGRAM, ops->icmp_protocol ); + if (s->socket < 0) { WARN( "Unable to create socket\n" ); - icmp_release_socket( data->s ); - free( data ); + pthread_mutex_unlock( &listen_lock ); + icmp_release_socket( s ); return STATUS_ACCESS_DENIED; } + setsockopt( s->socket, SOL_SOCKET, SO_REUSEADDR, &value, sizeof(value) ); #ifdef __linux__ if (ops->family == AF_INET) ops = &ipv4_linux_ping; - data->s->ping_socket = TRUE; + s->ping_socket = TRUE; #endif } - data->s->hop_limit = params->hop_limit; - data->s->ttl = params->ttl; - data->s->tos = params->tos; - src = (struct sockaddr *)&data->s->src_storage; - dst = (struct sockaddr *)&data->dst_storage; - data->s->src_len = SOCKADDR_INET_to_sockaddr( params->src, src, sizeof(data->s->src_storage) ); - data->dst_len = SOCKADDR_INET_to_sockaddr( params->dst, dst, sizeof(data->dst_storage) ); + s->hop_limit = params->hop_limit; + s->ttl = params->ttl; + s->tos = params->tos; + s->src_len = src_len; + s->src_storage = src_storage; - if (bind( data->s->socket, src, data->s->src_len )) + if (bind( s->socket, src, s->src_len )) { - icmp_data_free( data ); + pthread_mutex_unlock( &listen_lock ); + icmp_release_socket( s ); return STATUS_INVALID_ADDRESS_COMPONENT; } - data->s->ops = ops; - data->s->ops->set_socket_opts( data->s ); - pthread_mutex_lock( &listen_lock ); - list_add_tail( &socket_list, &data->s->entry ); - data->s->polling = TRUE; - write( socket_list_update_pipe[1], "x", 1 ); + s->ops = ops; + s->ops->set_socket_opts( s ); + list_add_tail( &socket_list, &s->entry ); + s->polling = TRUE; pthread_mutex_unlock( &listen_lock ); - - *icmp_data = data; - data->completion_event = params->completion_event; + write( socket_list_update_pipe[1], "x", 1 ); + *ret_socket = s; + TRACE( "created socket %p.\n", s ); return STATUS_SUCCESS; } @@ -926,12 +951,22 @@ NTSTATUS icmp_send_echo( void *args ) struct icmp_send_echo_params *params = args; struct icmp_hdr *icmp_hdr; /* this is the same for both ipv4 and ipv6 */ struct icmp_data *data; + struct sockaddr *dst; int ret, err; NTSTATUS status; - status = icmp_data_create( params, &data ); - if (status) return status; + data = calloc( 1, sizeof(*data) ); + if (!data) return STATUS_NO_MEMORY; + if ((status = icmp_get_socket( params, &data->s ))) + { + free( data ); + return status; + } + data->completion_event = params->completion_event; + + dst = (struct sockaddr *)&data->dst_storage; + data->dst_len = SOCKADDR_INET_to_sockaddr( params->dst, dst, sizeof(data->dst_storage) ); icmp_hdr = malloc( sizeof(*icmp_hdr) + params->request_size ); if (!icmp_hdr) @@ -943,6 +978,8 @@ NTSTATUS icmp_send_echo( void *args ) memcpy( icmp_hdr + 1, params->request, params->request_size ); icmp_hdr->checksum = data->s->ops->chksum( data, (BYTE *)icmp_hdr, sizeof(*icmp_hdr) + params->request_size ); + TRACE( "data %p, seq %d.\n", data, data->seq ); + pthread_mutex_lock( &listen_lock ); list_add_tail( &data->s->request_list, &data->entry ); data->polling = TRUE; -- GitLab https://gitlab.winehq.org/wine/wine/-/merge_requests/10954
From: Paul Gofman <pgofman@codeweavers.com> --- configure.ac | 1 + dlls/nsiproxy.sys/icmp_echo.c | 128 +++++++++++++++++++++++++++------- 2 files changed, 103 insertions(+), 26 deletions(-) diff --git a/configure.ac b/configure.ac index 5517c70ae94..9329ef83868 100644 --- a/configure.ac +++ b/configure.ac @@ -735,6 +735,7 @@ AC_CHECK_HEADERS(\ gettext-po.h \ link.h \ linux/cdrom.h \ + linux/errqueue.h \ linux/filter.h \ linux/fs.h \ linux/hdreg.h \ diff --git a/dlls/nsiproxy.sys/icmp_echo.c b/dlls/nsiproxy.sys/icmp_echo.c index d92b420715e..2fec9bffa8d 100644 --- a/dlls/nsiproxy.sys/icmp_echo.c +++ b/dlls/nsiproxy.sys/icmp_echo.c @@ -41,6 +41,11 @@ #include <netinet/ip.h> #endif +#ifdef HAVE_LINUX_ERRQUEUE_H +#include <linux/errqueue.h> +#define USE_ERRQUEUE +#endif + #include "ntstatus.h" #include "windef.h" #include "winbase.h" @@ -307,15 +312,16 @@ static void ipv4_linux_ping_set_socket_opts( struct icmp_socket *s ) setsockopt( s->socket, IPPROTO_IP, IP_RECVTTL, &val, sizeof(val) ); setsockopt( s->socket, IPPROTO_IP, IP_RECVTOS, &val, sizeof(val) ); + setsockopt( s->socket, IPPROTO_IP, IP_RECVERR, &val, sizeof(val) ); } #endif -static BOOL ipv4_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip_hdr_len, +static BOOL ipv4_parse_ip_hdr( struct msghdr *msg, int *recvd, int *ip_hdr_len, struct icmp_reply_ctx *ctx ) { struct ip_hdr *ip_hdr; - if (recvd < sizeof(*ip_hdr)) return FALSE; + if (*recvd < sizeof(*ip_hdr)) return FALSE; ip_hdr = msg->msg_iov[0].iov_base; if (ip_hdr->v_hl >> 4 != 4 || ip_hdr->protocol != IPPROTO_ICMP) return FALSE; *ip_hdr_len = (ip_hdr->v_hl & 0xf) << 2; @@ -330,7 +336,7 @@ static BOOL ipv4_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip_hdr_len, } #ifdef __linux__ -static BOOL ipv4_linux_ping_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip_hdr_len, +static BOOL ipv4_linux_ping_parse_ip_hdr( struct msghdr *msg, int *recvd, int *ip_hdr_len, struct icmp_reply_ctx *ctx ) { struct cmsghdr *cmsg; @@ -353,6 +359,25 @@ static BOOL ipv4_linux_ping_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip case IP_TOS: ctx->tos = *(BYTE *)CMSG_DATA( cmsg ); break; +#ifdef HAVE_LINUX_ERRQUEUE_H + case IP_RECVERR: + { + struct sock_extended_err *e; + struct icmp_hdr *icmp_h; + + e = (struct sock_extended_err *)CMSG_DATA( cmsg ); + icmp_h = msg->msg_iov[0].iov_base; + if (e->ee_origin != SO_EE_ORIGIN_ICMP) break; + if (*recvd + sizeof(*icmp_h) > ARRAY_SIZE(ctx->packet)) break; + memmove( ctx->packet + sizeof(*icmp_h), ctx->packet, *recvd ); + *recvd += sizeof(*icmp_h); + memset( icmp_h, 0, sizeof(*icmp_h) ); + icmp_h->type = e->ee_type; + icmp_h->code = e->ee_code; + TRACE( "got error packet type %d, code %d.\n", icmp_h->type, icmp_h->code ); + break; + } +#endif } } return TRUE; @@ -420,19 +445,22 @@ static int ipv4_parse_icmp_hdr( struct icmp_socket *s, struct icmp_hdr *icmp, in return -1; } + icmp_size -= sizeof(*icmp); if (s->ping_socket) { - /* Ping sockets only return echo reply through normal recvmsg() received data, should not get here. */ - return -1; + orig_ip_hdr_len = 0; + orig_icmp_hdr = icmp + 1; } - - /* All handled icmp replies have an 8-byte header followed by the original ip hdr. */ - if (icmp_size < sizeof(*icmp) + sizeof(*orig_ip_hdr)) return -1; - orig_ip_hdr = (struct ip_hdr *)(icmp + 1); - if (orig_ip_hdr->v_hl >> 4 != 4 || orig_ip_hdr->protocol != IPPROTO_ICMP) return -1; - orig_ip_hdr_len = (orig_ip_hdr->v_hl & 0xf) << 2; - if (icmp_size < sizeof(*icmp) + orig_ip_hdr_len + sizeof(*orig_icmp_hdr)) return -1; - orig_icmp_hdr = (const struct icmp_hdr *)((const BYTE *)orig_ip_hdr + orig_ip_hdr_len); + else + { + /* All handled icmp replies have an 8-byte header followed by the original ip hdr. */ + if (icmp_size < sizeof(*orig_ip_hdr)) return -1; + orig_ip_hdr = (struct ip_hdr *)(icmp + 1); + if (orig_ip_hdr->v_hl >> 4 != 4 || orig_ip_hdr->protocol != IPPROTO_ICMP) return -1; + orig_ip_hdr_len = (orig_ip_hdr->v_hl & 0xf) << 2; + orig_icmp_hdr = (const struct icmp_hdr *)((const BYTE *)orig_ip_hdr + orig_ip_hdr_len); + } + if (icmp_size < orig_ip_hdr_len + sizeof(*orig_icmp_hdr)) return -1; if (orig_icmp_hdr->type != ICMP4_ECHO_REQUEST || orig_icmp_hdr->code != 0) return -1; @@ -502,7 +530,7 @@ struct family_ops unsigned short (*chksum)( struct icmp_data *icmp_data, BYTE *data, unsigned int count ); int (*set_reply_ip_status)( IP_STATUS ip_status, unsigned int bits, void *out ); void (*set_socket_opts)( struct icmp_socket *s ); - BOOL (*parse_ip_hdr)( struct msghdr *msg, int recvd, int *ip_hdr_len, struct icmp_reply_ctx *ctx ); + BOOL (*parse_ip_hdr)( struct msghdr *msg, int *recvd, int *ip_hdr_len, struct icmp_reply_ctx *ctx ); int (*parse_icmp_hdr)( struct icmp_socket *s, struct icmp_hdr *icmp, int icmp_len, struct icmp_reply_ctx *ctx ); BOOL (*fill_reply)( struct icmp_get_reply_params *params, struct icmp_reply_ctx *ctx ); }; @@ -614,9 +642,17 @@ static void ipv6_set_socket_opts( struct icmp_socket *s ) setsockopt( s->socket, IPPROTO_IPV6, IPV6_UNICAST_HOPS, &val, sizeof(val) ); } #endif +#ifdef IP_RECVERR +{ + int val = 1; + + if (s->ping_socket) + setsockopt( s->socket, IPPROTO_IPV6, IPV6_RECVERR, &val, sizeof(val) ); +} +#endif } -static BOOL ipv6_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip_hdr_len, +static BOOL ipv6_parse_ip_hdr( struct msghdr *msg, int *recvd, int *ip_hdr_len, struct icmp_reply_ctx *ctx ) { *ip_hdr_len = 0; @@ -625,6 +661,34 @@ static BOOL ipv6_parse_ip_hdr( struct msghdr *msg, int recvd, int *ip_hdr_len, ctx->tos = 0; ctx->flags = 0; ctx->options_size = 0; + +#ifdef HAVE_LINUX_ERRQUEUE_H +{ + struct sock_extended_err *e; + struct icmp_hdr *icmp_h; + struct cmsghdr *cmsg; + + for (cmsg = CMSG_FIRSTHDR( msg ); cmsg; cmsg = CMSG_NXTHDR( msg, cmsg )) + { + if (cmsg->cmsg_level != IPPROTO_IPV6) continue; + switch (cmsg->cmsg_type) + { + case IPV6_RECVERR: + e = (struct sock_extended_err *)CMSG_DATA( cmsg ); + icmp_h = msg->msg_iov[0].iov_base; + if (e->ee_origin != SO_EE_ORIGIN_ICMP6) break; + if (*recvd + sizeof(*icmp_h) > ARRAY_SIZE(ctx->packet)) break; + TRACE( "got error packet.\n" ); + memmove( ctx->packet + sizeof(*icmp_h), ctx->packet, *recvd ); + *recvd += sizeof(*icmp_h); + memset( icmp_h, 0, sizeof(*icmp_h) ); + icmp_h->type = e->ee_type; + icmp_h->code = e->ee_code; + break; + } + } +} +#endif return TRUE; } @@ -682,18 +746,18 @@ static int ipv6_parse_icmp_hdr( struct icmp_socket *s, struct icmp_hdr *icmp, return -1; } - if (s->ping_socket) + icmp_size -= sizeof(*icmp); + if (s->ping_socket) orig_icmp_hdr = icmp + 1; + else { - /* Ping sockets only return echo reply through normal recvmsg() received data, should not get here. */ - return -1; + /* All handled icmp replies have an 8-byte header followed by the original ip hdr. */ + if (icmp_size < sizeof(*orig_ip_hdr)) return -1; + icmp_size -= sizeof(*orig_ip_hdr); + orig_ip_hdr = (struct ipv6_hdr *)(icmp + 1); + if ((orig_ip_hdr->v_prio >> 4) != 6 || orig_ip_hdr->next_hdr != IPPROTO_ICMPV6) return -1; + orig_icmp_hdr = (const struct icmp_hdr *)((const BYTE *)orig_ip_hdr + sizeof(*orig_ip_hdr)); } - - /* All handled icmp replies have an 8-byte header followed by the original ip hdr. */ - if (icmp_size < sizeof(*icmp) + sizeof(*orig_ip_hdr)) return -1; - orig_ip_hdr = (struct ipv6_hdr *)(icmp + 1); - if ((orig_ip_hdr->v_prio >> 4) != 6 || orig_ip_hdr->next_hdr != IPPROTO_ICMPV6) return -1; - if (icmp_size < sizeof(*icmp) + sizeof(*orig_ip_hdr) + sizeof(*orig_icmp_hdr)) return -1; - orig_icmp_hdr = (const struct icmp_hdr *)((const BYTE *)orig_ip_hdr + sizeof(*orig_ip_hdr)); + if (icmp_size < sizeof(*orig_icmp_hdr)) return -1; if (orig_icmp_hdr->type != ICMP6_ECHO_REQUEST || orig_icmp_hdr->code != 0) return -1; @@ -1081,8 +1145,20 @@ static void icmp_listen( void *args ) recvd = recvmsg( s->socket, &msg, 0 ); TRACE( "s %p, recvmsg() rets %d errno %d addr_len %d iovlen %d msg_flags %x\n", s, recvd, errno, msg.msg_namelen, (int)iov[0].iov_len, msg.msg_flags ); +#ifdef __linux__ + if (recvd < 0 && s->ping_socket && fds[i].revents & POLLERR) + { + msg.msg_name = &addr; + msg.msg_namelen = sizeof(addr); + msg.msg_control = cmsg_buf; + msg.msg_controllen = sizeof(cmsg_buf); + recvd = recvmsg( s->socket, &msg, MSG_ERRQUEUE ); + TRACE( "s %p, recvmsg(MSG_ERRQUEUE) rets %d errno %d addr_len %d iovlen %d msg_flags %x\n", + s, recvd, errno, msg.msg_namelen, (int)iov[0].iov_len, msg.msg_flags ); + } +#endif if (recvd < 0) goto skip; - if (!s->ops->parse_ip_hdr( &msg, recvd, &ip_hdr_len, &ctx )) goto skip; + if (!s->ops->parse_ip_hdr( &msg, &recvd, &ip_hdr_len, &ctx )) goto skip; if (recvd < ip_hdr_len + sizeof(*icmp_hdr)) goto skip; ctx.packet_size = recvd; icmp_hdr = (struct icmp_hdr *)(ctx.packet + ip_hdr_len); -- GitLab https://gitlab.winehq.org/wine/wine/-/merge_requests/10954
That is motivated by debugging SCUM game which could not connect to MP servers most of the time (while the game currently has some other blocking issue before that I believe this is generally useful). The game uses iphlpapi.IcmpSendEcho() for hundreds of servers. When the game fails to connect to (official) server that happens on network level, there are ICMP errors for TCP connection attempts (as seen through Wireshark). Those state that a packet was administratively filtered (while at another moment ping to that address from command line works). That doesn't happen (or doesn't happen that easily) on Windows or when allowing raw sockets for Wine with setcap(). Those errors come from some remote host (possibly intermediate). That seems to be triggered by a big amount of ICMP requests sent through IcmpSendEcho. The only solid difference I could find between ICMP over dgram vs Windows / Wine raw sockets is that ICMP packet id is constant on Windows (that is always 0001) and on Wine / raw sockets (nsiproxy process id). While with Linux ping sockets that id is set by the kernel to the UDP port number kernel assigned to the ping socket. Currently we create a separate socket for each ping request resulting in ever-changing id. Reusing the socket (s o it is bound to the same address and id is constant) seems to largely help the issue and it is possible to connect to servers (while refreshing server list a couple of times can trigger the issue again, while I could also reproduce that on Windows). While I don't know for sure, maybe that is some anti-DDOS filter which is sensitive to packet details and starts rejecting packets based on unusual contents. So the first 5 patches end up redesigning ICMP requests in nsiproxy.sys so that the socket is reused when possible. I believe that is overall beneficial: * besides more compatible ICMP packets, reusing sockets avoids allocating of UDP ports for each request which may result in hundreds of ports used concurrently; * removes extra unneeded threads, all the Unix side socket polling is done through one now and PE-side requests work with RegisterWaitForSingleObject; * Probably simplifies the thing a bit: remove 3 of 5 Unix calls related to ICMP, makes PE side code and IRP cancelling more straightforward. Then, currently ping sockets only work for echo replies while icmp error packets are not delivered for such sockets through normal recvmsg(). However, that can be retrieved using IP_RECVERR / IPV6_RECVERR msg. Ping sockets have special behaviour when MSG_ERRQUEUE is used and error packet is available: it will return original echo packet as the data received from recvmsg (that is needed to match the error packet to request and similar to what is available in raw icmp reply like that, just without reply IP packet header) and ICMP error type and code in 'struct sock_extended_err'. The last patch implements proper ICMP error reporting on ping sockets. Unfortunately I could not find anywhere reliable way to test that, whether these packets ever arrive depends on a lot of things in source and destination network configuration. I tested that locally by rejecting ICMP packets on [localhost](http://localhost) through ip\[6\]tables and verified how that works in game which receives those ICMP errors on various occasions (and it seems that delivering those insteading of timing out helps a bit additionally). -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10954#note_140723
participants (2)
-
Paul Gofman -
Paul Gofman (＠gofman)