Windows 10 [received support](https://devblogs.microsoft.com/commandline/af_unix-comes-to-windows/) for AF_UNIX sockets in Insider Build 17063. This merge request adds basic support for AF_UNIX sockets to ws2_32 and wineserver.
Of particular note is the difficulty in handling `sun_path`. Most of the functions that allow for translating Windows paths to Unix paths are not accessible from ws2_32. I considered the following options: * Pass the Windows path to wineserver and do the conversion there. * This is, as far as I can tell, not possible without major rearchitecting. wineserver does not have functions to translate Windows paths to Unix paths, for obvious reasons. * Obtain the current working directory of the requesting process and temporarily change directories to there. * This only handles relative paths and fails for absolute paths, UNC paths, etc. * Conditionally change directories based on whether the path is relative or not. * This is error-prone and wineserver does not have the requisite functions to do this cleanly.
I ultimately decided to pass the translated Unix path to wineserver, which changes directories to `dirname(path)`. It then provides `bind` and `connect` with `basename(path)`. This is not threadsafe, but wineserver is not (currently) multithreaded.
Abstract sockets are not fully supported by this patch, matching the behavior of Windows.
-- v47: ws2_32/tests: Add test for AF_UNIX sockets fix. server: Fix getsockname() and accept() on AF_UNIX sockets. server: Introduce error when attempting to create a SOCK_DGRAM AF_UNIX socket. server: Allow for deletion of socket files. ws2_32: Add support for AF_UNIX sockets. ntdll/unix: Add support for AF_UNIX sockets to multiple functions. ws2_32: Add afunix.h header.
From: Ally Sommers dropbear.sh@gmail.com
This header is needed for support of AF_UNIX sockets. --- include/Makefile.in | 1 + include/afunix.h | 38 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 39 insertions(+) create mode 100644 include/afunix.h
diff --git a/include/Makefile.in b/include/Makefile.in index 96f79518c2e..fa317f0ec5f 100644 --- a/include/Makefile.in +++ b/include/Makefile.in @@ -11,6 +11,7 @@ SOURCES = \ adshlp.h \ advpub.h \ af_irda.h \ + afunix.h \ amaudio.h \ amsi.idl \ amstream.idl \ diff --git a/include/afunix.h b/include/afunix.h new file mode 100644 index 00000000000..8118dda271b --- /dev/null +++ b/include/afunix.h @@ -0,0 +1,38 @@ +/* + * Copyright 2023 Ally Sommers + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * This library 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 + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA + */ + +#ifndef _WS2AFUNIX_ +#define _WS2AFUNIX_ + +#include "windef.h" +#include "ws2def.h" + +#ifdef USE_WS_PREFIX +# define WS(x) WS_##x +#else +# define WS(x) x +#endif + +#define UNIX_PATH_MAX 108 + +typedef struct WS(sockaddr_un) { + USHORT sun_family; + char sun_path[UNIX_PATH_MAX]; +} SOCKADDR_UN, *PSOCKADDR_UN; + +#endif /* _WS2AFUNIX_ */ \ No newline at end of file
From: Ally Sommers dropbear.sh@gmail.com
--- dlls/ntdll/unix/socket.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)
diff --git a/dlls/ntdll/unix/socket.c b/dlls/ntdll/unix/socket.c index e09bf34b07d..efd3904983a 100644 --- a/dlls/ntdll/unix/socket.c +++ b/dlls/ntdll/unix/socket.c @@ -67,6 +67,8 @@ # define HAS_IRDA #endif
+#include <sys/un.h> + #include "ntstatus.h" #define WIN32_NO_STATUS #include "windef.h" @@ -78,6 +80,7 @@ #include "ws2tcpip.h" #include "wsipx.h" #include "af_irda.h" +#include "afunix.h" #include "wine/afd.h"
#include "unix_private.h" @@ -106,6 +109,7 @@ union unix_sockaddr #ifdef HAS_IRDA struct sockaddr_irda irda; #endif + struct sockaddr_un un; };
struct async_recv_ioctl @@ -1953,6 +1957,14 @@ NTSTATUS sock_ioctl( HANDLE handle, HANDLE event, PIO_APC_ROUTINE apc, void *apc break; }
+ if (unix_addr.addr.sa_family == AF_UNIX) + { + FIXME( "getpeername called on AF_UNIX socket\n" ); + io->Information = sizeof(unix_addr.un); + status = STATUS_SUCCESS; + break; + } + len = sockaddr_from_unix( &unix_addr, out_buffer, out_size ); if (out_size < len) {
From: Ally Sommers dropbear.sh@gmail.com
This commit additionally modifies wineserver's sock_ioctl to handle the provided pathname by changing directories and then returning after the native call. This is NOT threadsafe, but wineserver is not multithreaded. --- dlls/ws2_32/socket.c | 86 +++++++++++++++++++++++++-- dlls/ws2_32/ws2_32_private.h | 14 +++++ server/sock.c | 111 ++++++++++++++++++++++++++++++++--- 3 files changed, 197 insertions(+), 14 deletions(-)
diff --git a/dlls/ws2_32/socket.c b/dlls/ws2_32/socket.c index 8ef7920e0fc..7dc6c2c912d 100644 --- a/dlls/ws2_32/socket.c +++ b/dlls/ws2_32/socket.c @@ -167,6 +167,19 @@ static const WSAPROTOCOL_INFOW supported_protocols[] = .dwMessageSize = UINT_MAX, .szProtocol = L"SPX II", }, + { + .dwServiceFlags1 = XP1_GUARANTEED_DELIVERY | XP1_GUARANTEED_ORDER | XP1_IFS_HANDLES, + .dwProviderFlags = PFL_MATCHES_PROTOCOL_ZERO, + .ProviderId = {0xa00943d9, 0x9c2e, 0x4633, {0x9b, 0x59, 0x00, 0x57, 0xa3, 0x16, 0x09, 0x94}}, + .dwCatalogEntryId = 1007, + .ProtocolChain.ChainLen = 1, + .iVersion = 2, + .iAddressFamily = AF_UNIX, + .iMaxSockAddr = sizeof(struct sockaddr_un), + .iMinSockAddr = offsetof(struct sockaddr_un, sun_path), + .iSocketType = SOCK_STREAM, + .szProtocol = L"AF_UNIX", + }, };
DECLARE_CRITICAL_SECTION(cs_socket_list); @@ -226,6 +239,11 @@ const char *debugstr_sockaddr( const struct sockaddr *a ) addr, ((const SOCKADDR_IRDA *)a)->irdaServiceName); } + case AF_UNIX: + { + return wine_dbg_sprintf("{ family AF_UNIX, path %s }", + ((const SOCKADDR_UN *)a)->sun_path); + } default: return wine_dbg_sprintf("{ family %d }", a->sa_family); } @@ -1106,6 +1124,10 @@ int WINAPI bind( SOCKET s, const struct sockaddr *addr, int len ) HANDLE sync_event; NTSTATUS status;
+ const int bind_len = len; + char *unix_path = NULL; + int unix_path_len = 0; + TRACE( "socket %#Ix, addr %s\n", s, debugstr_sockaddr(addr) );
if (!addr) @@ -1148,6 +1170,14 @@ int WINAPI bind( SOCKET s, const struct sockaddr *addr, int len ) } break;
+ case AF_UNIX: + if (len < offsetof(struct sockaddr_un, sun_path)) + { + SetLastError( WSAEFAULT ); + return -1; + } + break; + default: FIXME( "unknown protocol %u\n", addr->sa_family ); SetLastError( WSAEAFNOSUPPORT ); @@ -1156,8 +1186,30 @@ int WINAPI bind( SOCKET s, const struct sockaddr *addr, int len )
if (!(sync_event = get_sync_event())) return -1;
- params = malloc( sizeof(int) + len ); - ret_addr = malloc( len ); + if (addr->sa_family == AF_UNIX && *addr->sa_data) + { + struct sockaddr_un sun = { 0 }; + WCHAR *sun_pathW; + memcpy(&sun, addr, len); + if (strlen( sun.sun_path )) + { + sun_pathW = strdupAtoW( sun.sun_path ); + unix_path = wine_get_unix_file_name( sun_pathW ); + free( sun_pathW ); + if (!unix_path) + return SOCKET_ERROR; + } + else + { + unix_path = malloc(1); + *unix_path = '\0'; + } + unix_path_len = strlen( unix_path ) + 1; + len = sizeof(sun); + } + + params = malloc( sizeof(int) + len + unix_path_len ); + ret_addr = malloc( len + unix_path_len ); if (!params || !ret_addr) { free( params ); @@ -1166,10 +1218,15 @@ int WINAPI bind( SOCKET s, const struct sockaddr *addr, int len ) return -1; } params->unknown = 0; - memcpy( ¶ms->addr, addr, len ); + memcpy( ¶ms->addr, addr, bind_len ); + if (unix_path) + { + ((char *)¶ms->addr)[bind_len] = '\0'; + memcpy( (char *)¶ms->addr + len, unix_path, unix_path_len ); + }
status = NtDeviceIoControlFile( (HANDLE)s, sync_event, NULL, NULL, &io, IOCTL_AFD_BIND, - params, sizeof(int) + len, ret_addr, len ); + params, sizeof(int) + len + unix_path_len, ret_addr, len + unix_path_len ); if (status == STATUS_PENDING) { if (WaitForSingleObject( sync_event, INFINITE ) == WAIT_FAILED) @@ -1182,6 +1239,7 @@ int WINAPI bind( SOCKET s, const struct sockaddr *addr, int len )
free( params ); free( ret_addr ); + free( unix_path );
SetLastError( NtStatusToWSAError( status ) ); return status ? -1 : 0; @@ -1222,11 +1280,24 @@ int WINAPI connect( SOCKET s, const struct sockaddr *addr, int len ) HANDLE sync_event; NTSTATUS status;
+ char *unix_path = NULL; + int unix_path_len = 0; + TRACE( "socket %#Ix, addr %s, len %d\n", s, debugstr_sockaddr(addr), len );
if (!(sync_event = get_sync_event())) return -1;
- if (!(params = malloc( sizeof(*params) + len ))) + if (addr->sa_family == AF_UNIX && *addr->sa_data) + { + WCHAR *sun_pathW = strdupAtoW(addr->sa_data); + unix_path = wine_get_unix_file_name(sun_pathW); + free(sun_pathW); + if (!unix_path) + return SOCKET_ERROR; + unix_path_len = strlen(unix_path) + 1; + } + + if (!(params = malloc( sizeof(*params) + len + unix_path_len ))) { SetLastError( ERROR_NOT_ENOUGH_MEMORY ); return -1; @@ -1234,10 +1305,13 @@ int WINAPI connect( SOCKET s, const struct sockaddr *addr, int len ) params->addr_len = len; params->synchronous = TRUE; memcpy( params + 1, addr, len ); + if (unix_path_len) + memcpy( (char *)(params + 1) + len, unix_path, unix_path_len );
status = NtDeviceIoControlFile( (HANDLE)s, sync_event, NULL, NULL, &io, IOCTL_AFD_WINE_CONNECT, - params, sizeof(*params) + len, NULL, 0 ); + params, sizeof(*params) + len + unix_path_len, NULL, 0 ); free( params ); + free( unix_path ); if (status == STATUS_PENDING) { if (wait_event_alertable( sync_event ) == WAIT_FAILED) return -1; diff --git a/dlls/ws2_32/ws2_32_private.h b/dlls/ws2_32/ws2_32_private.h index 6ebcd6837ae..481be1a170e 100644 --- a/dlls/ws2_32/ws2_32_private.h +++ b/dlls/ws2_32/ws2_32_private.h @@ -44,6 +44,7 @@ #include "mstcpip.h" #include "af_irda.h" #include "winnt.h" +#include "afunix.h" #define USE_WC_PREFIX /* For CMSG_DATA */ #include "iphlpapi.h" #include "ip2string.h" @@ -71,6 +72,19 @@ static inline char *strdupWtoA( const WCHAR *str ) return ret; }
+static inline WCHAR *strdupAtoW( const char *str ) +{ + WCHAR *ret = NULL; + if (str) + { + DWORD len = MultiByteToWideChar(CP_ACP, 0, str, -1, NULL, 0); + if ((ret = malloc( len * sizeof(WCHAR) ))) + MultiByteToWideChar(CP_ACP, 0, str, -1, ret, len); + } + return ret; +} + + static const char magic_loopback_addr[] = {127, 12, 34, 56};
const char *debugstr_sockaddr( const struct sockaddr *addr ) DECLSPEC_HIDDEN; diff --git a/server/sock.c b/server/sock.c index ce95b7f2998..ecd918b6074 100644 --- a/server/sock.c +++ b/server/sock.c @@ -53,6 +53,7 @@ #include <time.h> #include <unistd.h> #include <limits.h> +#include <libgen.h> #ifdef HAVE_LINUX_FILTER_H # include <linux/filter.h> #endif @@ -84,6 +85,8 @@ # define HAS_IRDA #endif
+#include <sys/un.h> + #include "ntstatus.h" #define WIN32_NO_STATUS #include "windef.h" @@ -94,6 +97,7 @@ #include "ws2tcpip.h" #include "wsipx.h" #include "af_irda.h" +#include "afunix.h" #include "wine/afd.h" #include "wine/rbtree.h"
@@ -117,6 +121,7 @@ union win_sockaddr struct WS_sockaddr_in6 in6; struct WS_sockaddr_ipx ipx; SOCKADDR_IRDA irda; + struct WS_sockaddr_un un; };
union unix_sockaddr @@ -130,6 +135,7 @@ union unix_sockaddr #ifdef HAS_IRDA struct sockaddr_irda irda; #endif + struct sockaddr_un un; };
static struct list poll_list = LIST_INIT( poll_list ); @@ -673,6 +679,9 @@ static socklen_t get_unix_sockaddr_any( union unix_sockaddr *uaddr, int ws_famil uaddr->irda.sir_family = AF_IRDA; return sizeof(uaddr->irda); #endif + case WS_AF_UNIX: + uaddr->un.sun_family = AF_UNIX; + return sizeof(uaddr->un); default: return 0; } @@ -1757,6 +1766,7 @@ static int get_unix_family( int family ) #ifdef AF_IRDA case WS_AF_IRDA: return AF_IRDA; #endif + case WS_AF_UNIX: return AF_UNIX; case WS_AF_UNSPEC: return AF_UNSPEC; default: return -1; } @@ -2527,8 +2537,13 @@ static void sock_ioctl( struct fd *fd, ioctl_code_t code, struct async *async )
if (listen( unix_fd, params->backlog ) < 0) { - set_error( sock_get_ntstatus( errno ) ); - return; + /* Due to the way we handle the Windows AF_UNIX bind edge case, we also need to + * ignore listen's error. */ + if (!(errno == EINVAL && sock->family == WS_AF_UNIX && !*sock->addr.un.sun_path)) + { + set_error( sock_get_ntstatus( errno ) ); + return; + } }
sock->state = SOCK_LISTENING; @@ -2598,7 +2613,36 @@ static void sock_ioctl( struct fd *fd, ioctl_code_t code, struct async *async ) break; }
- unix_len = sockaddr_to_unix( addr, params->addr_len, &unix_addr ); + if (sock->family == WS_AF_UNIX) + { + if (*((struct WS_sockaddr_un *)&addr)->sun_path) + { + char *unix_path = (char *)(params + 1) + params->addr_len; + char unix_path_copy[PATH_MAX]; + send_len -= strlen( unix_path ) + 1; + strcpy(unix_path_copy, unix_path); + if (chdir( dirname( unix_path_copy ) ) == -1) + { + set_error( sock_get_ntstatus( errno ) ); + return; + } + + unix_len = sizeof(unix_addr.un); + unix_addr.un.sun_family = AF_UNIX; + memcpy( unix_addr.un.sun_path, basename( unix_path ), sizeof(unix_addr.un.sun_path) ); + } + else + { + /* Contrary to documentation, Windows does not currently support abstract Unix + * sockets. connect() throws WSAEINVAL if sun_family is AF_UNIX and sun_path + * begins with '\0', even though bind() will succeed. */ + set_win32_error( WSAEINVAL ); + return; + } + } + else + unix_len = sockaddr_to_unix( addr, params->addr_len, &unix_addr ); + if (!unix_len) { set_error( STATUS_INVALID_ADDRESS ); @@ -2619,6 +2663,9 @@ static void sock_ioctl( struct fd *fd, ioctl_code_t code, struct async *async ) ret = connect( unix_fd, &unix_addr.addr, unix_len ); }
+ if (sock->family == WS_AF_UNIX && *((struct WS_sockaddr_un *)&addr)->sun_path) + fchdir(server_dir_fd); + if (ret < 0 && errno != EINPROGRESS) { set_error( sock_get_ntstatus( errno ) ); @@ -2882,7 +2929,29 @@ static void sock_ioctl( struct fd *fd, ioctl_code_t code, struct async *async ) return; }
- unix_len = sockaddr_to_unix( ¶ms->addr, in_size - sizeof(int), &unix_addr ); + if (sock->family == WS_AF_UNIX) + { + if (*params->addr.sa_data) + { + char *unix_path = (char *)¶ms->addr + sizeof(unix_addr.un); + char unix_path_copy[PATH_MAX]; + strcpy( unix_path_copy, unix_path ); + if (chdir( dirname( unix_path_copy ) ) == -1) + { + set_error( sock_get_ntstatus( errno ) ); + return; + } + + memcpy( unix_addr.un.sun_path, basename( unix_path ), sizeof(unix_addr.un.sun_path) ); + } + else + memset(unix_addr.un.sun_path, 0, sizeof(unix_addr.un.sun_path)); + unix_addr.un.sun_family = AF_UNIX; + unix_len = sizeof(unix_addr.un); + } + else + unix_len = sockaddr_to_unix( ¶ms->addr, in_size - sizeof(int), &unix_addr ); + if (!unix_len) { set_error( STATUS_INVALID_ADDRESS ); @@ -2926,8 +2995,16 @@ static void sock_ioctl( struct fd *fd, ioctl_code_t code, struct async *async ) if (errno == EADDRINUSE && sock->reuseaddr) errno = EACCES;
- set_error( sock_get_ntstatus( errno ) ); - return; + /* Windows' AF_UNIX implementation has an edge case allowing for a socket to bind to + * an empty path. Linux doesn't, so it throws EINVAL. We check for this situation + * here and avoid early-exiting if it's the case. */ + if (!(errno == EINVAL && sock->family == WS_AF_UNIX && !*params->addr.sa_data)) + { + set_error( sock_get_ntstatus( errno ) ); + if (sock->family == WS_AF_UNIX && *params->addr.sa_data) + fchdir(server_dir_fd); + return; + } }
sock->bound = 1; @@ -2939,13 +3016,23 @@ static void sock_ioctl( struct fd *fd, ioctl_code_t code, struct async *async ) * actual unix address */ if (bind_addr.addr.sa_family == AF_INET) bind_addr.in.sin_addr = unix_addr.in.sin_addr; - sock->addr_len = sockaddr_from_unix( &bind_addr, &sock->addr.addr, sizeof(sock->addr) ); + if (bind_addr.addr.sa_family == AF_UNIX) + { + sock->addr.un.sun_family = WS_AF_UNIX; + memcpy(sock->addr.un.sun_path, params->addr.sa_data, sizeof(sock->addr.un.sun_path)); + sock->addr_len = sizeof(sock->addr.un); + } + else + sock->addr_len = sockaddr_from_unix( &bind_addr, &sock->addr.addr, sizeof(sock->addr) ); }
update_addr_usage( sock, &bind_addr, v6only );
if (get_reply_max_size() >= sock->addr_len) set_reply_data( &sock->addr, sock->addr_len ); + + if (sock->family == WS_AF_UNIX && *params->addr.sa_data) + fchdir(server_dir_fd); return; }
@@ -2962,7 +3049,15 @@ static void sock_ioctl( struct fd *fd, ioctl_code_t code, struct async *async ) return; }
- set_reply_data( &sock->addr, sock->addr_len ); + if (sock->family == WS_AF_UNIX) + { + if (*sock->addr.un.sun_path) + set_reply_data( &sock->addr, sizeof(sock->addr.un.sun_family) + strlen(sock->addr.un.sun_path) + 1 ); + else + set_reply_data( &sock->addr, sizeof(sock->addr.un) ); + } + else + set_reply_data( &sock->addr, sock->addr_len ); return;
case IOCTL_AFD_WINE_DEFER:
From: Ally Sommers dropbear.sh@gmail.com
Deleting the socket file is a common pattern with AF_UNIX sockets, and is analogous to unbinding. --- server/fd.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+)
diff --git a/server/fd.c b/server/fd.c index 0b0e91ebfbb..98d5ecdbcc3 100644 --- a/server/fd.c +++ b/server/fd.c @@ -1971,6 +1971,27 @@ struct fd *open_fd( struct fd *root, const char *name, struct unicode_str nt_nam fd->unix_fd = open( name, O_RDONLY | (flags & ~(O_TRUNC | O_CREAT | O_EXCL)), *mode ); }
+ /* POSIX requires that open(2) throws EOPNOTSUPP when `path` is a Unix + * socket. *BSD throws EOPNOTSUPP in this case and the additional case of + * O_SHLOCK or O_EXLOCK being passed when `path` resides on a filesystem + * without lock support. + * + * Contrary to POSIX, Linux returns ENXIO in this case, so we also check + * that error code here. */ + if ((errno == EOPNOTSUPP || errno == ENXIO) && + !stat( name, &st ) && + S_ISSOCK(st.st_mode) && + (access & DELETE)) + { + if (unlink( name )) + { + file_set_error(); + goto error; + } + else + fd->unix_fd = open( name, O_CREAT | O_RDWR, S_IRWXU ); + } + if (fd->unix_fd == -1) { /* check for trailing slash on file path */
From: Ally Sommers dropbear.sh@gmail.com
--- server/sock.c | 6 ++++++ 1 file changed, 6 insertions(+)
diff --git a/server/sock.c b/server/sock.c index ecd918b6074..feca245cf12 100644 --- a/server/sock.c +++ b/server/sock.c @@ -1859,6 +1859,12 @@ static int init_socket( struct sock *sock, int family, int type, int protocol ) return -1; }
+ if (unix_family == AF_UNIX && unix_type == SOCK_DGRAM) + { + set_win32_error(WSAEAFNOSUPPORT); + return -1; + } + sockfd = socket( unix_family, unix_type, unix_protocol );
#ifdef linux
From: Ally Sommers dropbear.sh@gmail.com
--- server/sock.c | 33 ++++++++++++++++++++++++++++++--- 1 file changed, 30 insertions(+), 3 deletions(-)
diff --git a/server/sock.c b/server/sock.c index feca245cf12..b685002789a 100644 --- a/server/sock.c +++ b/server/sock.c @@ -2036,7 +2036,19 @@ static struct sock *accept_socket( struct sock *sock ) } unix_len = sizeof(unix_addr); if (!getsockname( acceptfd, &unix_addr.addr, &unix_len )) - acceptsock->addr_len = sockaddr_from_unix( &unix_addr, &acceptsock->addr.addr, sizeof(acceptsock->addr) ); + { + if (sock->family == WS_AF_UNIX) + { + acceptsock->addr_len = sock->addr_len; + acceptsock->addr.un = sock->addr.un; + } + else + { + acceptsock->addr_len = sockaddr_from_unix( &unix_addr, + &acceptsock->addr.addr, + sizeof(acceptsock->addr) ); + } + } } clear_error(); sock->pending_events &= ~AFD_POLL_ACCEPT; @@ -2092,7 +2104,19 @@ static int accept_into_socket( struct sock *sock, struct sock *acceptsock )
unix_len = sizeof(unix_addr); if (!getsockname( get_unix_fd( newfd ), &unix_addr.addr, &unix_len )) - acceptsock->addr_len = sockaddr_from_unix( &unix_addr, &acceptsock->addr.addr, sizeof(acceptsock->addr) ); + { + if (sock->family == WS_AF_UNIX) + { + acceptsock->addr_len = sock->addr_len; + acceptsock->addr.un = sock->addr.un; + } + else + { + acceptsock->addr_len = sockaddr_from_unix( &unix_addr, + &acceptsock->addr.addr, + sizeof(acceptsock->addr) ); + } + }
clear_error(); sock->pending_events &= ~AFD_POLL_ACCEPT; @@ -2683,7 +2707,10 @@ static void sock_ioctl( struct fd *fd, ioctl_code_t code, struct async *async )
unix_len = sizeof(unix_addr); if (!getsockname( unix_fd, &unix_addr.addr, &unix_len )) - sock->addr_len = sockaddr_from_unix( &unix_addr, &sock->addr.addr, sizeof(sock->addr) ); + { + if (sock->family != WS_AF_UNIX) + sock->addr_len = sockaddr_from_unix( &unix_addr, &sock->addr.addr, sizeof(sock->addr) ); + } sock->bound = 1;
if (!ret)
From: Ally Sommers dropbear.sh@gmail.com
--- dlls/ws2_32/tests/sock.c | 292 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 292 insertions(+)
diff --git a/dlls/ws2_32/tests/sock.c b/dlls/ws2_32/tests/sock.c index e53d9991e56..af781427453 100644 --- a/dlls/ws2_32/tests/sock.c +++ b/dlls/ws2_32/tests/sock.c @@ -29,6 +29,7 @@ #include <iphlpapi.h> #include <ws2tcpip.h> #include <wsipx.h> +#include <afunix.h> #include <wsnwlink.h> #include <mswsock.h> #include <mstcpip.h> @@ -13820,6 +13821,296 @@ static void test_tcp_sendto_recvfrom(void) closesocket(server); }
+static void test_afunix(void) +{ + SOCKET listener, client, server = 0; + SOCKADDR_UN addr = { AF_UNIX, "test_afunix.sock" }; + char serverBuf[] = "ws2_32/AF_UNIX socket test"; + char clientBuf[sizeof(serverBuf)] = { 0 }; + + char paths[4][sizeof(addr.sun_path)] = { "./tmp.sock", "../tmp.sock" }; + char dosPath[sizeof(addr.sun_path)]; + WCHAR dosWidePath[sizeof(addr.sun_path)]; + UNICODE_STRING ntPath = { 0 }; + SOCKADDR_UN outAddr = { 0 }; + int outAddrSize = sizeof(outAddr); + SOCKADDR_UN truncatedAddr = { 0 }; + ULONG zero = 0; + ULONG one = 1; + int ret; + + /* Test connection and send/recv */ + listener = socket(AF_UNIX, SOCK_STREAM, 0); + if (listener == INVALID_SOCKET && GetLastError() == WSAEAFNOSUPPORT) + { + win_skip("AF_UNIX sockets are unsupported, skipping...\n"); + return; + } + + ok(listener != INVALID_SOCKET, "Could not create Unix socket: %lu\n", + GetLastError()); + + ret = bind(listener, (SOCKADDR *)&addr, 0); + ok(ret && GetLastError() == WSAEFAULT, "Incorrect error: %lu\n", GetLastError()); + ret = bind(listener, (SOCKADDR *)&addr, 2); + ok(!ret, "Could not bind Unix socket: %lu\n", GetLastError()); + ret = listen(listener, 1); + ok(!ret, "Could not listen on Unix socket: %lu\n", GetLastError()); + closesocket(listener); + + listener = socket(AF_UNIX, SOCK_STREAM, 0); + ok(listener != INVALID_SOCKET, "Could not create Unix socket: %lu\n", + GetLastError()); + ret = bind(listener, (SOCKADDR *)&addr, 3); + ok(!ret, "Could not bind Unix socket: %lu\n", GetLastError()); + + memcpy(&truncatedAddr, &addr, 3); + ret = getsockname(listener, (SOCKADDR *)&outAddr, &outAddrSize); + ok(!ret, "Could not get info on Unix socket: %lu\n", GetLastError()); + ok(!memcmp(truncatedAddr.sun_path, outAddr.sun_path, sizeof(addr.sun_path)), + "getsockname returned incorrect path '%s' for provided path '%s'\n", + outAddr.sun_path, + truncatedAddr.sun_path); + ok(outAddrSize == sizeof(outAddr.sun_family) + strlen(outAddr.sun_path) + 1, + "getsockname returned incorrect size '%d' for provided path '%s'\n", + outAddrSize, + truncatedAddr.sun_path); + closesocket(listener); + ret = DeleteFileA(truncatedAddr.sun_path); + ok(ret, "DeleteFileA on socket file failed: %lu\n", GetLastError()); + ok(GetFileAttributesA(truncatedAddr.sun_path) == INVALID_FILE_ATTRIBUTES && + GetLastError() == ERROR_FILE_NOT_FOUND, + "Failed to delete socket file at path '%s'\n", + truncatedAddr.sun_path); + + listener = socket(AF_UNIX, SOCK_STREAM, 0); + ok(listener != INVALID_SOCKET, "Could not create Unix socket: %lu\n", + GetLastError()); + + ret = bind(listener, (SOCKADDR *)&addr, sizeof(SOCKADDR_UN)); + ok(!ret, "Could not bind Unix socket: %lu\n", GetLastError()); + + ret = listen(listener, 1); + ok(!ret, "Could not listen on Unix socket: %lu\n", GetLastError()); + + client = socket(AF_UNIX, SOCK_STREAM, 0); + ok(client != INVALID_SOCKET, "Failed to create second Unix socket: %lu\n", + GetLastError()); + + ret = ioctlsocket(client, FIONBIO, &one); + ok(!ret, "Could not set AF_UNIX socket to nonblocking: %lu; skipping connection\n", GetLastError()); + if (!ret) + { + ret = connect(client, (SOCKADDR *)&addr, sizeof(addr)); + ok(!ret || (ret == SOCKET_ERROR && GetLastError() == WSAEWOULDBLOCK), + "Error when connecting to Unix socket: %lu\n", + GetLastError()); + server = accept(listener, NULL, NULL); + ok(server != INVALID_SOCKET, "Could not accept Unix socket connection: %lu\n", + GetLastError()); + ret = ioctlsocket(client, FIONBIO, &zero); + ok(!ret, "Could not set AF_UNIX socket to blocking: %lu\n", GetLastError()); + } + + ret = send(server, serverBuf, sizeof(serverBuf), 0); + ok(ret == sizeof(serverBuf), "Incorrect return value from send: %d\n", ret); + ret = recv(client, clientBuf, sizeof(serverBuf), 0); + ok(ret == sizeof(serverBuf), "Incorrect return value from recv: %d\n", ret); + ok(!memcmp(serverBuf, clientBuf, sizeof(serverBuf)), "Data mismatch over Unix socket\n"); + + memset(clientBuf, 0, sizeof(clientBuf)); + + ret = sendto(server, serverBuf, sizeof(serverBuf), 0, NULL, 0); + ok(ret == sizeof(serverBuf), "Incorrect return value from sendto: %d\n", ret); + ret = recvfrom(client, clientBuf, sizeof(serverBuf), 0, NULL, 0); + ok(ret == sizeof(serverBuf), "Incorrect return value from recvfrom: %d\n", ret); + ok(!memcmp(serverBuf, clientBuf, sizeof(serverBuf)), "Data mismatch over Unix socket\n"); + + memset(serverBuf, 0, sizeof(serverBuf)); + + ret = send(client, clientBuf, sizeof(clientBuf), 0); + ok(ret == sizeof(clientBuf), "Incorrect return value from send: %d\n", ret); + ret = recv(server, serverBuf, sizeof(clientBuf), 0); + ok(ret == sizeof(serverBuf), "Incorrect return value from recv: %d\n", ret); + ok(!memcmp(serverBuf, clientBuf, sizeof(clientBuf)), "Data mismatch over Unix socket\n"); + + memset(serverBuf, 0, sizeof(serverBuf)); + + ret = sendto(client, clientBuf, sizeof(clientBuf), 0, NULL, 0); + ok(ret == sizeof(clientBuf), "Incorrect return value from sendto: %d\n", ret); + ret = recvfrom(server, serverBuf, sizeof(clientBuf), 0, NULL, 0); + ok(ret == sizeof(serverBuf), "Incorrect return value from recvfrom: %d\n", ret); + ok(!memcmp(serverBuf, clientBuf, sizeof(clientBuf)), "Data mismatch over Unix socket\n"); + + closesocket(listener); + closesocket(client); + closesocket(server); + + /* Test socket file deletion */ + ret = DeleteFileA("test_afunix.sock"); + ok(ret, "DeleteFileA on socket file failed: %lu\n", GetLastError()); + ok(GetFileAttributesA("test_afunix.sock") == INVALID_FILE_ATTRIBUTES && + GetLastError() == ERROR_FILE_NOT_FOUND, + "Failed to delete socket file at path '%s'\n", + addr.sun_path); + + /* Test failure modes */ + listener = socket(AF_UNIX, SOCK_STREAM, 0); + ok(listener != INVALID_SOCKET, "Could not create Unix socket: %lu\n", + GetLastError()); + ret = bind(listener, (SOCKADDR *)&addr, sizeof(SOCKADDR_UN)); + ok(!ret, "Could not bind Unix socket to path '%s': %lu\n", addr.sun_path, GetLastError()); + closesocket(listener); + listener = socket(AF_UNIX, SOCK_STREAM, 0); + ok(listener != INVALID_SOCKET, "Could not create Unix socket: %lu\n", + GetLastError()); + ret = bind(listener, (SOCKADDR *)&addr, sizeof(SOCKADDR_UN)); + ok(ret && GetLastError() == WSAEADDRINUSE, + "Bound Unix socket to path '%s' despite existing socket file: %lu\n", + addr.sun_path, + GetLastError()); + closesocket(listener); + ret = DeleteFileA(addr.sun_path); + ok(ret, "DeleteFileA on socket file failed: %lu\n", GetLastError()); + ok(GetFileAttributesA("test_afunix.sock") == INVALID_FILE_ATTRIBUTES && + GetLastError() == ERROR_FILE_NOT_FOUND, + "Failed to delete socket file at path '%s'\n", + addr.sun_path); + + /* Test different path types (relative, NT, etc.) */ + GetTempPathA(sizeof(paths[0]) - 1, paths[2]); + strcat(paths[2], "tmp.sock"); + MultiByteToWideChar(CP_ACP, 0, paths[2], -1, dosWidePath, sizeof(dosPath)); + RtlDosPathNameToNtPathName_U(dosWidePath, &ntPath, NULL, NULL); + RtlUnicodeToMultiByteN(paths[3], sizeof(addr.sun_path) - 1, NULL, ntPath.Buffer, ntPath.Length); + + for (int i = 0; i < sizeof(paths) / sizeof(paths[0]); i++) + { + memcpy(addr.sun_path, paths[i], sizeof(paths[i])); + + listener = socket(AF_UNIX, SOCK_STREAM, 0); + ok(listener != INVALID_SOCKET, "Could not create Unix socket: %lu\n", + GetLastError()); + + ret = bind(listener, (SOCKADDR *)&addr, sizeof(SOCKADDR_UN)); + ok(!ret, "Could not bind Unix socket to path '%s': %lu\n", addr.sun_path, GetLastError()); + + ret = listen(listener, 1); + ok(!ret, "Could not listen on Unix socket: %lu\n", GetLastError()); + + client = socket(AF_UNIX, SOCK_STREAM, 0); + ok(client != INVALID_SOCKET, "Failed to create second Unix socket: %lu\n", + GetLastError()); + + ret = ioctlsocket(client, FIONBIO, &one); + ok(!ret, "Could not set AF_UNIX socket to nonblocking: %lu; skipping connection\n", GetLastError()); + if (!ret) + { + ret = connect(client, (SOCKADDR *)&addr, sizeof(addr)); + ok(!ret || (ret == SOCKET_ERROR && GetLastError() == WSAEWOULDBLOCK), + "Error when connecting to Unix socket: %lu\n", + GetLastError()); + server = accept(listener, NULL, NULL); + ok(server != INVALID_SOCKET, "Could not accept Unix socket connection: %lu\n", + GetLastError()); + ret = ioctlsocket(client, FIONBIO, &zero); + ok(!ret, "Could not set AF_UNIX socket to blocking: %lu\n", GetLastError()); + } + + memset(&outAddr, 0, sizeof(outAddr)); + outAddrSize = sizeof(outAddr); + ret = getsockname(listener, (SOCKADDR *)&outAddr, &outAddrSize); + ok(!ret, "Could not get info on Unix socket: %lu\n", GetLastError()); + ok(!memcmp(addr.sun_path, outAddr.sun_path, sizeof(addr.sun_path)), + "getsockname returned incorrect path '%s' for provided path '%s'\n", + outAddr.sun_path, + addr.sun_path); + ok(outAddrSize == sizeof(outAddr.sun_family) + strlen(outAddr.sun_path) + 1, + "getsockname returned incorrect size '%d' for provided path '%s'\n", + outAddrSize, + addr.sun_path); + + memset(&outAddr, 0, sizeof(outAddr)); + outAddrSize = sizeof(outAddr); + ret = getsockname(client, (SOCKADDR *)&outAddr, &outAddrSize); + ok(!ret, "Could not get info on Unix socket: %lu\n", GetLastError()); + ok(!memcmp((char[108]){0}, outAddr.sun_path, sizeof(addr.sun_path)), + "getsockname returned incorrect path '%s' for provided path '%s'\n", + outAddr.sun_path, + addr.sun_path); + ok(outAddrSize == sizeof(outAddr), + "getsockname returned incorrect size '%d' for provided path '%s'\n", + outAddrSize, + addr.sun_path); + + memset(&outAddr, 0, sizeof(outAddr)); + outAddrSize = sizeof(outAddr); + ret = getsockname(server, (SOCKADDR *)&outAddr, &outAddrSize); + ok(!ret, "Could not get info on Unix socket: %lu\n", GetLastError()); + ok(!memcmp(addr.sun_path, outAddr.sun_path, sizeof(addr.sun_path)), + "getsockname returned incorrect path '%s' for provided path '%s'\n", + outAddr.sun_path, + addr.sun_path); + ok(outAddrSize == sizeof(outAddr.sun_family) + strlen(outAddr.sun_path) + 1, + "getsockname returned incorrect size '%d' for provided path '%s'\n", + outAddrSize, + addr.sun_path); + + memset(&outAddr, 0, sizeof(outAddr)); + outAddrSize = sizeof(outAddr); + ret = getpeername(listener, (SOCKADDR *)&outAddr, &outAddrSize); + ok(ret == -1, "Got info on Unix socket: %lu\n", GetLastError()); + ok(GetLastError() == WSAENOTCONN, + "Incorrect error returned from getpeername on Unix socket: %ld\n", + GetLastError()); + ok(!memcmp((char[108]){0}, outAddr.sun_path, sizeof(addr.sun_path)), + "getpeername returned incorrect path '%s' for provided path '%s'\n", + outAddr.sun_path, + addr.sun_path); + ok(outAddrSize == sizeof(outAddr), + "getpeername returned incorrect size '%d' for provided path '%s'\n", + outAddrSize, + addr.sun_path); + + memset(&outAddr, 0, sizeof(outAddr)); + outAddrSize = sizeof(outAddr); + ret = getpeername(client, (SOCKADDR *)&outAddr, &outAddrSize); + ok(!ret, "Could not get info on Unix socket: %lu\n", GetLastError()); + todo_wine ok(!memcmp(addr.sun_path, outAddr.sun_path, sizeof(addr.sun_path)), + "getpeername returned incorrect path '%s' for provided path '%s'\n", + outAddr.sun_path, + addr.sun_path); + ok(outAddrSize == sizeof(outAddr), + "getpeername returned incorrect size '%d' for provided path '%s'\n", + outAddrSize, + addr.sun_path); + + memset(&outAddr, 0, sizeof(outAddr)); + outAddrSize = sizeof(outAddr); + ret = getpeername(server, (SOCKADDR *)&outAddr, &outAddrSize); + ok(!ret, "Could not get info on Unix socket: %lu\n", GetLastError()); + ok(!memcmp((char[108]){0}, outAddr.sun_path, sizeof(addr.sun_path)), + "getpeername returned incorrect path '%s' for provided path '%s'\n", + outAddr.sun_path, + addr.sun_path); + ok(outAddrSize == sizeof(outAddr), + "getpeername returned incorrect size '%d' for provided path '%s'\n", + outAddrSize, + addr.sun_path); + + closesocket(listener); + closesocket(client); + closesocket(server); + + ret = DeleteFileA(addr.sun_path); + ok(ret, "DeleteFileA on socket file failed: %lu\n", GetLastError()); + ok(GetFileAttributesA(addr.sun_path) == INVALID_FILE_ATTRIBUTES && + GetLastError() == ERROR_FILE_NOT_FOUND, + "Failed to delete socket file at path '%s'\n", + addr.sun_path); + } +} + START_TEST( sock ) { int i; @@ -13901,6 +14192,7 @@ START_TEST( sock ) test_icmp(); test_connect_udp(); test_tcp_sendto_recvfrom(); + test_afunix();
/* There is apparently an obscure interaction between this test and * test_WSAGetOverlappedResult().
On Sun Aug 13 14:26:17 2023 +0000, Loïc Rebmeister wrote:
Any news? @dropbear has been working on maintaining the patch for multiples weeks now without any interaction. It would be great if there was some feedback at some point to what will happens to this patch. The longer it takes for to review patches, the more of @dropbear time will be used on maintaining old patch instead of trying to make new one to improve WineHQ. (I don't say patches have to be merged, just to not waste precious wine contributors time so Wine can move forward faster)
* Of the last 1000 merged commits, only about 11 have focused on ws2_32 (4 that didn't only add/modify tests). Maintaining the patch is thus not hard. I haven't had to make any code changes other than resolving merge conflicts since it's been approved; all of the tests still pass. * Since this patch took so long to get into a mergeable state, it's probably not as high up on the list of patches to be merged, although I obviously don't know Mr. Julliard's workflow. * I'm not able to devote too much time to Wine at the moment, so my time isn't being "wasted", per se, not that it would be even if I could given the low maintenance requirement of this patch.
On Sun Aug 13 14:26:33 2023 +0000, Ally Sommers wrote:
- Of the last 1000 merged commits, only about 11 have focused on ws2_32
(4 that didn't only add/modify tests). Maintaining the patch is thus not hard. I haven't had to make any code changes other than resolving merge conflicts since it's been approved; all of the tests still pass.
- Since this patch took so long to get into a mergeable state, it's
probably not as high up on the list of patches to be merged, although I obviously don't know Mr. Julliard's workflow.
- I'm not able to devote too much time to Wine at the moment, so my time
isn't being "wasted", per se, not that it would be even if I could given the low maintenance requirement of this patch.
Well, it's frustrating to see no feedback on patches for so long.
I completely understand that sometimes: - it can take some time to respond to a patch. - patches still needs more work. - patches should not be merged. - life is more important than code.
But seeing no feedback on the status of this patch for so long is frustrating to see.
Alexandre has been out and not reviewing nontrivial patches for the past couple of weeks. That said, this patch series has been stalled for longer than that, and if I had to guess what the problem is, I'd say the most worrying part is the socket deletion logic. I don't know how to improve that, though.
The deletion logic is indeed ugly, it should be possible to find a better way. As a first step I'd suggest moving it to a separate MR.
Moving the socket file deletion to another MR would preclude having tests for it, because we'd have no way of actually creating a socket file to delete (this is done by `bind`). It would also cause almost all of the tests in this MR to fail, because the tests reuse the same socket path and we'd have no way to delete the existing socket files from previous tests.
I can experiment with different implementations for deletion; I admittedly don't yet fully understand the codepath that file deletion takes and there may well be a cleaner way of doing this.
-
Alexandre Julliard (＠julliard) -
Ally Sommers
-
Ally Sommers (＠dropbear)
-
Loïc Rebmeister (＠Fox2Code)
-
Zebediah Figura (＠zfigura)