March 2022 - Wine-Devel - list.winehq.org

[website] German translation for release 7.2
by Julian Rüger March 3, 2022

March 3, 2022

Signed-off-by: Julian Rüger <jr98(a)gmx.net>

1 0

[website] German translation for WWN issue 411
by Julian Rüger March 3, 2022

March 3, 2022

Signed-off-by: Julian Rüger <jr98(a)gmx.net>

1 0

[website] German translation for WWN issue 410
by Julian Rüger March 3, 2022

March 3, 2022

Signed-off-by: Julian Rüger <jr98(a)gmx.net>

1 0

[website] German translation for release 7.1
by Julian Rüger March 3, 2022

March 3, 2022

Signed-off-by: Julian Rüger <jr98(a)gmx.net>

1 0

[PATCH] wined3d: Fix typo in WARN() message.
by Zebediah Figura March 3, 2022

March 3, 2022

From: Matteo Bruni <mbruni(a)codeweavers.com> Signed-off-by: Matteo Bruni <mbruni(a)codeweavers.com> Signed-off-by: Zebediah Figura <zfigura(a)codeweavers.com> --- dlls/wined3d/state.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dlls/wined3d/state.c b/dlls/wined3d/state.c index 1bab18b8547..32e4a3e75a4 100644 --- a/dlls/wined3d/state.c +++ b/dlls/wined3d/state.c @@ -4217,7 +4217,7 @@ static void state_cs_uav_binding(struct wined3d_context *context, static void state_uav_warn(struct wined3d_context *context, const struct wined3d_state *state, DWORD state_id) { - WARN("ARB_image_load_store is not supported by OpenGL implementation.\n"); + WARN("ARB_shader_image_load_store is not supported by this OpenGL implementation.\n"); } static void state_so(struct wined3d_context *context, const struct wined3d_state *state, DWORD state_id) -- 2.35.1

2 1

[PATCH v4 0/8] Fix sock_send reordering issue (#52401)
by Jinoh Kang March 3, 2022

March 3, 2022

This fixes the send part of the issue #52401 [1]: "Improper synchronization in sock_recv/sock_send leads to arbitrary reordering of completion of I/O requests", which was initially reported (outside Bugzilla) by Dongwan Kim [2]. Changelog: v1 -> v2: - drop patch "server: Compact struct set_async_direct_result_reply." - add async and fd arguments to async_initial_status_callback - recv_socket_initial_callback: pass OOB flag via private instead of implementing two separate functions for OOB and non-OOB - detect short write in send_socket_initial_callback - preserve the behaviour of returning success if we had a short write and the socket is nonblocking and force_async is unset v2 -> v3: - fix typo in comment v3 -> v4: - fix typo in commit message - address feedback - drop patch "server: Defer postprocessing until after setting initial status in recv_socket handler." [1] https://bugs.winehq.org/show_bug.cgi?id=52401 [2] https://www.winehq.org/pipermail/wine-devel/2021-May/186454.html Jinoh Kang (8): server: Actually set initial status in set_async_direct_result handler. server: Ensure initial status is set in async_set_result(). server: Add async_initial_status_callback. server: Defer postprocessing until after setting initial status in send_socket handler. server: Add mark_pending field to set_async_direct_result request. server: Attempt to complete I/O request immediately in send_socket. ntdll: Don't call try_send before server call in sock_send. server: Replace redundant send_socket status fields with force_async boolean field. dlls/ntdll/unix/socket.c | 102 +++++++++++++++++++++++++-------- dlls/ntdll/unix/sync.c | 9 +-- dlls/ntdll/unix/unix_private.h | 2 +- dlls/ws2_32/tests/sock.c | 14 ++--- server/async.c | 60 ++++++++++++++----- server/file.h | 2 + server/protocol.def | 6 +- server/sock.c | 92 ++++++++++++++++------------- 8 files changed, 194 insertions(+), 93 deletions(-) Interdiff: diff --git a/server/async.c b/server/async.c index ded4f8392d2..90fede50371 100644 --- a/server/async.c +++ b/server/async.c @@ -326,12 +326,8 @@ void async_set_initial_status_callback( struct async *async, async_initial_statu * the initial status may be STATUS_PENDING */ void async_set_initial_status( struct async *async, unsigned int status ) { - assert( async->unknown_status ); - if (!async->terminated) - { - set_async_initial_status( async, status ); - async->unknown_status = 0; - } + set_async_initial_status( async, status ); + async->unknown_status = 0; } void set_async_pending( struct async *async ) @@ -509,6 +505,8 @@ void async_set_result( struct object *obj, unsigned int status, apc_param_t tota assert( async->terminated ); /* it must have been woken up if we get a result */ + if (async->unknown_status) async_set_initial_status( async, status ); + if (async->alerted && status == STATUS_PENDING) /* restart it */ { async->terminated = 0; @@ -797,11 +795,6 @@ DECL_HANDLER(set_async_direct_result) return; } - /* Set iosb information field for async_initial_status_callback */ - async->iosb->result = information; - set_async_initial_status( async, status ); - async->unknown_status = 0; - if (status == STATUS_PENDING) { async->direct_result = 0; @@ -812,6 +805,9 @@ DECL_HANDLER(set_async_direct_result) async->pending = 1; } + /* Set iosb information field for async_initial_status_callback */ + async->iosb->result = information; + /* if the I/O has completed successfully (or unsuccessfully, and * async->pending is set), the client would have already set the IOSB. * therefore, we can do async_set_result() directly and let the client skip diff --git a/server/sock.c b/server/sock.c index 1112708369d..c742bd687a1 100644 --- a/server/sock.c +++ b/server/sock.c @@ -3393,16 +3393,6 @@ struct object *create_socket_device( struct object *root, const struct unicode_s return create_named_object( root, &socket_device_ops, name, attr, sd ); } -static void recv_socket_initial_callback( void *private, struct async *async, struct fd *fd, unsigned int status ) -{ - struct sock *sock = get_fd_user( fd ); - int is_oob = (int)(long)private; - - sock->pending_events &= ~(is_oob ? AFD_POLL_OOB : AFD_POLL_READ); - sock->reported_events &= ~(is_oob ? AFD_POLL_OOB : AFD_POLL_READ); - sock_reselect( sock ); -} - DECL_HANDLER(recv_socket) { struct sock *sock = (struct sock *)get_handle_obj( current->process, req->async.handle, 0, &sock_ops ); @@ -3440,18 +3430,22 @@ DECL_HANDLER(recv_socket) if (status == STATUS_PENDING && !req->force_async && sock->nonblocking) status = STATUS_DEVICE_NOT_READY; + sock->pending_events &= ~(req->oob ? AFD_POLL_OOB : AFD_POLL_READ); + sock->reported_events &= ~(req->oob ? AFD_POLL_OOB : AFD_POLL_READ); + if ((async = create_request_async( fd, get_fd_comp_flags( fd ), &req->async ))) { set_error( status ); - async_set_initial_status_callback( async, recv_socket_initial_callback, (void *)(long)req->oob ); - if (timeout) async_set_timeout( async, timeout, STATUS_IO_TIMEOUT ); if (status == STATUS_PENDING || status == STATUS_ALERTED) queue_async( &sock->read_q, async ); + /* always reselect; we changed reported_events above */ + sock_reselect( sock ); + reply->wait = async_handoff( async, NULL, 0 ); reply->options = get_fd_options( fd ); reply->nonblocking = sock->nonblocking; -- 2.34.1

2 1

[PATCH v4 8/8] server: Replace redundant send_socket status fields with force_async boolean field.
by Jinoh Kang March 3, 2022

March 3, 2022

The 'status' field of send_socket_request is always either STATUS_PENDING or STATUS_DEVICE_NOT_READY, and the 'total' field is always zero. Replace the 'status' field with 'force_async' boolean field, and get rid of the 'total' field entirely. Also, clean up the send_socket handler code a bit. Signed-off-by: Jinoh Kang <jinoh.kang.kr(a)gmail.com> --- Notes: v1 -> v2: - adjust patch for new field "data_size" in send_socket_request v2 -> v3: no changes v3 -> v4: no changes dlls/ntdll/unix/socket.c | 8 ++------ server/protocol.def | 5 ++--- server/sock.c | 40 ++++++++-------------------------------- 3 files changed, 12 insertions(+), 41 deletions(-) diff --git a/dlls/ntdll/unix/socket.c b/dlls/ntdll/unix/socket.c index ad03907523d..11ce652fed5 100644 --- a/dlls/ntdll/unix/socket.c +++ b/dlls/ntdll/unix/socket.c @@ -921,13 +921,10 @@ static NTSTATUS sock_send( HANDLE handle, HANDLE event, PIO_APC_ROUTINE apc, voi data_size += len; } - status = force_async ? STATUS_PENDING : STATUS_DEVICE_NOT_READY; - SERVER_START_REQ( send_socket ) { req->data_size = data_size; - req->status = status; - req->total = 0; + req->force_async = force_async; req->async = server_async( handle, &async->io, event, apc, apc_user, iosb_client_ptr(io) ); status = wine_server_call( req ); wait_handle = wine_server_ptr_handle( reply->wait ); @@ -1131,8 +1128,7 @@ static NTSTATUS sock_transmit( HANDLE handle, HANDLE event, PIO_APC_ROUTINE apc, SERVER_START_REQ( send_socket ) { req->data_size = async->head_len + async->file_len + async->tail_len; - req->status = STATUS_PENDING; - req->total = 0; + req->force_async = 1; req->async = server_async( handle, &async->io, event, apc, apc_user, iosb_client_ptr(io) ); status = wine_server_call( req ); wait_handle = wine_server_ptr_handle( reply->wait ); diff --git a/server/protocol.def b/server/protocol.def index 334abb66a77..5d8038d58b9 100644 --- a/server/protocol.def +++ b/server/protocol.def @@ -1461,10 +1461,9 @@ enum server_fd_type /* Perform a send on a socket */ @REQ(send_socket) - data_size_t data_size; /* total number of bytes to send (>= total) */ + data_size_t data_size; /* total number of bytes to send */ async_data_t async; /* async I/O parameters */ - unsigned int status; /* status of initial call */ - unsigned int total; /* number of bytes already sent */ + int force_async; /* Force asynchronous mode? */ @REPLY obj_handle_t wait; /* handle to wait on for blocking send */ unsigned int options; /* device open options */ diff --git a/server/sock.c b/server/sock.c index e899695bb60..c742bd687a1 100644 --- a/server/sock.c +++ b/server/sock.c @@ -3492,7 +3492,7 @@ static void send_socket_initial_callback( void *private, struct async *async, st DECL_HANDLER(send_socket) { struct sock *sock = (struct sock *)get_handle_obj( current->process, req->async.handle, 0, &sock_ops ); - unsigned int status = req->status; + unsigned int status = STATUS_PENDING; timeout_t timeout = 0; struct async *async; struct fd *fd; @@ -3500,32 +3500,11 @@ DECL_HANDLER(send_socket) if (!sock) return; fd = sock->fd; - /* If we had a short write and the socket is nonblocking (and the client is - * not trying to force the operation to be asynchronous), return success. - * Windows actually refuses to send any data in this case, and returns - * EWOULDBLOCK, but we have no way of doing that. */ - if (status == STATUS_DEVICE_NOT_READY && req->total && sock->nonblocking) - status = STATUS_SUCCESS; + if (!req->force_async && !sock->nonblocking && is_fd_overlapped( fd )) + timeout = (timeout_t)sock->sndtimeo * -10000; - /* send() returned EWOULDBLOCK or a short write, i.e. cannot send all data yet */ - if (status == STATUS_DEVICE_NOT_READY && !sock->nonblocking) - { - /* Set a timeout on the async if necessary. - * - * We want to do this *only* if the client gave us STATUS_DEVICE_NOT_READY. - * If the client gave us STATUS_PENDING, it expects the async to always - * block (it was triggered by WSASend*() with a valid OVERLAPPED - * structure) and for the timeout not to be respected. */ - if (is_fd_overlapped( fd )) - timeout = (timeout_t)sock->sndtimeo * -10000; - - status = STATUS_PENDING; - } - - if ((status == STATUS_PENDING || status == STATUS_DEVICE_NOT_READY) && sock->wr_shutdown) - status = STATUS_PIPE_DISCONNECTED; - - if ((status == STATUS_PENDING || status == STATUS_DEVICE_NOT_READY) && !async_queued( &sock->write_q )) + if (sock->wr_shutdown) status = STATUS_PIPE_DISCONNECTED; + else if (!async_queued( &sock->write_q )) { /* If write_q is not empty, we cannot really tell if the already queued * asyncs will not consume all available space; if there's no space @@ -3544,14 +3523,11 @@ DECL_HANDLER(send_socket) } } + if (status == STATUS_PENDING && !req->force_async && sock->nonblocking) + status = STATUS_DEVICE_NOT_READY; + if ((async = create_request_async( fd, get_fd_comp_flags( fd ), &req->async ))) { - if (status == STATUS_SUCCESS) - { - struct iosb *iosb = async_get_iosb( async ); - iosb->result = req->total; - release_object( iosb ); - } set_error( status ); async_set_initial_status_callback( async, send_socket_initial_callback, (void *)(unsigned long)req->data_size ); -- 2.34.1

1 0

[PATCH v4 7/8] ntdll: Don't call try_send before server call in sock_send.
by Jinoh Kang March 3, 2022

March 3, 2022

Otherwise, try_send() call from sock_send() may race against try_send() call from async_send_proc(), shuffling the packet order. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=52401 Signed-off-by: Jinoh Kang <jinoh.kang.kr(a)gmail.com> --- Notes: v1 -> v2: no changes v2 -> v3: no changes v3 -> v4: no changes dlls/ntdll/unix/socket.c | 13 ++----------- dlls/ws2_32/tests/sock.c | 14 +++++++------- 2 files changed, 9 insertions(+), 18 deletions(-) diff --git a/dlls/ntdll/unix/socket.c b/dlls/ntdll/unix/socket.c index 23014ceceaf..ad03907523d 100644 --- a/dlls/ntdll/unix/socket.c +++ b/dlls/ntdll/unix/socket.c @@ -921,22 +921,13 @@ static NTSTATUS sock_send( HANDLE handle, HANDLE event, PIO_APC_ROUTINE apc, voi data_size += len; } - status = try_send( fd, async ); - - if (status != STATUS_SUCCESS && status != STATUS_DEVICE_NOT_READY) - { - release_fileio( &async->io ); - return status; - } - - if (status == STATUS_DEVICE_NOT_READY && force_async) - status = STATUS_PENDING; + status = force_async ? STATUS_PENDING : STATUS_DEVICE_NOT_READY; SERVER_START_REQ( send_socket ) { req->data_size = data_size; req->status = status; - req->total = async->sent_len; + req->total = 0; req->async = server_async( handle, &async->io, event, apc, apc_user, iosb_client_ptr(io) ); status = wine_server_call( req ); wait_handle = wine_server_ptr_handle( reply->wait ); diff --git a/dlls/ws2_32/tests/sock.c b/dlls/ws2_32/tests/sock.c index 6b6267dc153..caf76868f89 100644 --- a/dlls/ws2_32/tests/sock.c +++ b/dlls/ws2_32/tests/sock.c @@ -7799,7 +7799,7 @@ static void test_shutdown(void) WSASetLastError(0xdeadbeef); ret = send(client, "test", 5, 0); ok(ret == -1, "got %d\n", ret); - todo_wine ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); + ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); ret = recv(server, buffer, sizeof(buffer), 0); ok(!ret, "got %d\n", ret); @@ -7846,7 +7846,7 @@ static void test_shutdown(void) WSASetLastError(0xdeadbeef); ret = send(server, "test", 5, 0); ok(ret == -1, "got %d\n", ret); - todo_wine ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); + ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); addrlen = sizeof(addr); ret = getpeername(client, (struct sockaddr *)&addr, &addrlen); @@ -7896,7 +7896,7 @@ static void test_shutdown(void) WSASetLastError(0xdeadbeef); ret = send(client, "test", 5, 0); ok(ret == -1, "got %d\n", ret); - todo_wine ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); + ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); ret = recv(server, buffer, sizeof(buffer), 0); ok(!ret, "got %d\n", ret); @@ -7914,7 +7914,7 @@ static void test_shutdown(void) WSASetLastError(0xdeadbeef); ret = send(server, "test", 5, 0); ok(ret == -1, "got %d\n", ret); - todo_wine ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); + ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); addrlen = sizeof(addr); ret = getpeername(client, (struct sockaddr *)&addr, &addrlen); @@ -8045,7 +8045,7 @@ static void test_shutdown(void) WSASetLastError(0xdeadbeef); ret = sendto(client, "test", 5, 0, (struct sockaddr *)&server_addr, sizeof(server_addr)); ok(ret == -1, "got %d\n", ret); - todo_wine ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); + ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); closesocket(client); closesocket(server); @@ -8126,7 +8126,7 @@ static void test_DisconnectEx(void) WSASetLastError(0xdeadbeef); ret = send(client, "test", 5, 0); ok(ret == -1, "expected failure\n"); - todo_wine ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); + ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); ret = recv(server, buffer, sizeof(buffer), 0); ok(!ret, "got %d\n", ret); @@ -8180,7 +8180,7 @@ static void test_DisconnectEx(void) WSASetLastError(0xdeadbeef); ret = send(client, "test", 5, 0); ok(ret == -1, "expected failure\n"); - todo_wine ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); + ok(WSAGetLastError() == WSAESHUTDOWN, "got error %u\n", WSAGetLastError()); ret = recv(server, buffer, sizeof(buffer), 0); ok(!ret, "got %d\n", ret); -- 2.34.1

1 0

[PATCH v4 6/8] server: Attempt to complete I/O request immediately in send_socket.
by Jinoh Kang March 3, 2022

March 3, 2022

Make send_socket alert the async immediately if poll() call detects that there are incoming data in the socket, bypassing the wineserver's main polling loop. For sock_transmit, we always mark the async as pending and set the IOSB (unless async allocation has failed). Signed-off-by: Jinoh Kang <jinoh.kang.kr(a)gmail.com> --- Notes: v1 -> v2: - retain the behaviour of returning success if we had a short write and the socket is nonblocking and force_async is unset v2 -> v3: fix typo in comment v3 -> v4: no changes dlls/ntdll/unix/socket.c | 72 ++++++++++++++++++++++++++++++++++------ server/protocol.def | 1 + server/sock.c | 22 +++++++++++- 3 files changed, 83 insertions(+), 12 deletions(-) diff --git a/dlls/ntdll/unix/socket.c b/dlls/ntdll/unix/socket.c index f24215b66a9..23014ceceaf 100644 --- a/dlls/ntdll/unix/socket.c +++ b/dlls/ntdll/unix/socket.c @@ -868,12 +868,14 @@ static NTSTATUS sock_send( HANDLE handle, HANDLE event, PIO_APC_ROUTINE apc, voi const struct WS_sockaddr *addr, unsigned int addr_len, int unix_flags, int force_async ) { struct async_send_ioctl *async; + ULONG_PTR information; HANDLE wait_handle; DWORD async_size; NTSTATUS status; unsigned int i; ULONG options; data_size_t data_size; + BOOL nonblocking, alerted; async_size = offsetof( struct async_send_ioctl, iov[count] ); @@ -939,16 +941,38 @@ static NTSTATUS sock_send( HANDLE handle, HANDLE event, PIO_APC_ROUTINE apc, voi status = wine_server_call( req ); wait_handle = wine_server_ptr_handle( reply->wait ); options = reply->options; - if ((!NT_ERROR(status) || wait_handle) && status != STATUS_PENDING) + nonblocking = reply->nonblocking; + } + SERVER_END_REQ; + + alerted = status == STATUS_ALERTED; + if (alerted) + { + status = try_send( fd, async ); + if (status == STATUS_DEVICE_NOT_READY && (force_async || !nonblocking)) + status = STATUS_PENDING; + + /* If we had a short write and the socket is nonblocking (and we are + * not trying to force the operation to be asynchronous), return + * success. Windows actually refuses to send any data in this case, + * and returns EWOULDBLOCK, but we have no way of doing that. */ + if (status == STATUS_DEVICE_NOT_READY && async->sent_len) + status = STATUS_SUCCESS; + } + + if (status != STATUS_PENDING) + { + information = async->sent_len; + if (!NT_ERROR(status) || (wait_handle && !alerted)) { io->Status = status; - io->Information = async->sent_len; + io->Information = information; } + release_fileio( &async->io ); } - SERVER_END_REQ; - - if (status != STATUS_PENDING) release_fileio( &async->io ); + else information = 0; + if (alerted) set_async_direct_result( &wait_handle, status, information, FALSE ); if (wait_handle) status = wait_async( wait_handle, options & FILE_SYNCHRONOUS_IO_ALERT ); return status; } @@ -1069,7 +1093,9 @@ static NTSTATUS sock_transmit( HANDLE handle, HANDLE event, PIO_APC_ROUTINE apc, socklen_t addr_len; HANDLE wait_handle; NTSTATUS status; + ULONG_PTR information; ULONG options; + BOOL alerted; addr_len = sizeof(addr); if (getpeername( fd, &addr.addr, &addr_len ) != 0) @@ -1120,16 +1146,40 @@ static NTSTATUS sock_transmit( HANDLE handle, HANDLE event, PIO_APC_ROUTINE apc, status = wine_server_call( req ); wait_handle = wine_server_ptr_handle( reply->wait ); options = reply->options; - /* In theory we'd fill the iosb here, as above in sock_send(), but it's - * actually currently impossible to get STATUS_SUCCESS. The server will - * either return STATUS_PENDING or an error code, and in neither case - * should the iosb be filled. */ - if (!status) FIXME( "Unhandled success status." ); } SERVER_END_REQ; - if (status != STATUS_PENDING) release_fileio( &async->io ); + alerted = status == STATUS_ALERTED; + if (alerted) + { + status = try_transmit( fd, file_fd, async ); + if (status == STATUS_DEVICE_NOT_READY) + status = STATUS_PENDING; + } + if (status != STATUS_PENDING) + { + information = async->head_cursor + async->file_cursor + async->tail_cursor; + if (!NT_ERROR(status) || wait_handle) + { + io->Status = status; + io->Information = information; + } + release_fileio( &async->io ); + } + else information = 0; + + if (alerted) + { + set_async_direct_result( &wait_handle, status, information, TRUE ); + if (!(options & (FILE_SYNCHRONOUS_IO_ALERT | FILE_SYNCHRONOUS_IO_NONALERT))) + { + /* Pretend we always do async I/O. The client can always retrieve + * the actual I/O status via the IO_STATUS_BLOCK. + */ + status = STATUS_PENDING; + } + } if (wait_handle) status = wait_async( wait_handle, options & FILE_SYNCHRONOUS_IO_ALERT ); return status; } diff --git a/server/protocol.def b/server/protocol.def index 5eb63db3091..334abb66a77 100644 --- a/server/protocol.def +++ b/server/protocol.def @@ -1468,6 +1468,7 @@ enum server_fd_type @REPLY obj_handle_t wait; /* handle to wait on for blocking send */ unsigned int options; /* device open options */ + int nonblocking; /* is socket non-blocking? */ @END diff --git a/server/sock.c b/server/sock.c index 0e386e3537c..e899695bb60 100644 --- a/server/sock.c +++ b/server/sock.c @@ -3525,6 +3525,25 @@ DECL_HANDLER(send_socket) if ((status == STATUS_PENDING || status == STATUS_DEVICE_NOT_READY) && sock->wr_shutdown) status = STATUS_PIPE_DISCONNECTED; + if ((status == STATUS_PENDING || status == STATUS_DEVICE_NOT_READY) && !async_queued( &sock->write_q )) + { + /* If write_q is not empty, we cannot really tell if the already queued + * asyncs will not consume all available space; if there's no space + * available, the current request won't be immediately satiable. + */ + struct pollfd pollfd; + pollfd.fd = get_unix_fd( sock->fd ); + pollfd.events = POLLOUT; + pollfd.revents = 0; + if (poll(&pollfd, 1, 0) >= 0 && pollfd.revents) + { + /* Give the client opportunity to complete synchronously. + * If it turns out that the I/O request is not actually immediately satiable, + * the client may then choose to re-queue the async (with STATUS_PENDING). */ + status = STATUS_ALERTED; + } + } + if ((async = create_request_async( fd, get_fd_comp_flags( fd ), &req->async ))) { if (status == STATUS_SUCCESS) @@ -3540,11 +3559,12 @@ DECL_HANDLER(send_socket) if (timeout) async_set_timeout( async, timeout, STATUS_IO_TIMEOUT ); - if (status == STATUS_PENDING) + if (status == STATUS_PENDING || status == STATUS_ALERTED) queue_async( &sock->write_q, async ); reply->wait = async_handoff( async, NULL, 0 ); reply->options = get_fd_options( fd ); + reply->nonblocking = sock->nonblocking; release_object( async ); } release_object( sock ); -- 2.34.1

1 0

[PATCH v4 5/8] server: Add mark_pending field to set_async_direct_result request.
by Jinoh Kang March 3, 2022

March 3, 2022

The client can set mark_pending to indicate that the full-blown I/O completion mechanism shall be triggered (asynchronous completion) even if the status indicates failure. Signed-off-by: Jinoh Kang <jinoh.kang.kr(a)gmail.com> --- Notes: v1 -> v2: no changes v2 -> v3: no changes v3 -> v4: no changes dlls/ntdll/unix/socket.c | 2 +- dlls/ntdll/unix/sync.c | 9 +++++---- dlls/ntdll/unix/unix_private.h | 2 +- server/async.c | 11 ++++++++--- server/protocol.def | 1 + 5 files changed, 16 insertions(+), 9 deletions(-) diff --git a/dlls/ntdll/unix/socket.c b/dlls/ntdll/unix/socket.c index d04ebb7f874..f24215b66a9 100644 --- a/dlls/ntdll/unix/socket.c +++ b/dlls/ntdll/unix/socket.c @@ -767,7 +767,7 @@ static NTSTATUS sock_recv( HANDLE handle, HANDLE event, PIO_APC_ROUTINE apc, voi release_fileio( &async->io ); } - if (alerted) set_async_direct_result( &wait_handle, status, information ); + if (alerted) set_async_direct_result( &wait_handle, status, information, FALSE ); if (wait_handle) status = wait_async( wait_handle, options & FILE_SYNCHRONOUS_IO_ALERT ); return status; } diff --git a/dlls/ntdll/unix/sync.c b/dlls/ntdll/unix/sync.c index ee25dfd0099..0786454dad2 100644 --- a/dlls/ntdll/unix/sync.c +++ b/dlls/ntdll/unix/sync.c @@ -2514,7 +2514,7 @@ NTSTATUS WINAPI NtWaitForAlertByThreadId( const void *address, const LARGE_INTEG /* Notify direct completion of async and close the wait handle if it is no longer needed. * This function is a no-op (returns status as-is) if the supplied handle is NULL. */ -void set_async_direct_result( HANDLE *optional_handle, NTSTATUS status, ULONG_PTR information ) +void set_async_direct_result( HANDLE *optional_handle, NTSTATUS status, ULONG_PTR information, BOOL mark_pending ) { NTSTATUS ret; @@ -2522,9 +2522,10 @@ void set_async_direct_result( HANDLE *optional_handle, NTSTATUS status, ULONG_PT SERVER_START_REQ( set_async_direct_result ) { - req->handle = wine_server_obj_handle( *optional_handle ); - req->status = status; - req->information = information; + req->handle = wine_server_obj_handle( *optional_handle ); + req->status = status; + req->information = information; + req->mark_pending = mark_pending; ret = wine_server_call( req ); if (ret == STATUS_SUCCESS) *optional_handle = wine_server_ptr_handle( reply->handle ); diff --git a/dlls/ntdll/unix/unix_private.h b/dlls/ntdll/unix/unix_private.h index b1595c792c0..b1a834efba5 100644 --- a/dlls/ntdll/unix/unix_private.h +++ b/dlls/ntdll/unix/unix_private.h @@ -274,7 +274,7 @@ extern NTSTATUS get_device_info( int fd, struct _FILE_FS_DEVICE_INFORMATION *inf extern void init_files(void) DECLSPEC_HIDDEN; extern void init_cpu_info(void) DECLSPEC_HIDDEN; extern void add_completion( HANDLE handle, ULONG_PTR value, NTSTATUS status, ULONG info, BOOL async ) DECLSPEC_HIDDEN; -extern void set_async_direct_result( HANDLE *optional_handle, NTSTATUS status, ULONG_PTR information ); +extern void set_async_direct_result( HANDLE *optional_handle, NTSTATUS status, ULONG_PTR information, BOOL mark_pending ); extern void dbg_init(void) DECLSPEC_HIDDEN; diff --git a/server/async.c b/server/async.c index 986e8a1ce24..90fede50371 100644 --- a/server/async.c +++ b/server/async.c @@ -800,13 +800,18 @@ DECL_HANDLER(set_async_direct_result) async->direct_result = 0; async->pending = 1; } + else if (req->mark_pending) + { + async->pending = 1; + } /* Set iosb information field for async_initial_status_callback */ async->iosb->result = information; - /* if the I/O has completed successfully, the client would have already - * set the IOSB. therefore, we can skip waiting on wait_handle and do - * async_set_result() directly. + /* if the I/O has completed successfully (or unsuccessfully, and + * async->pending is set), the client would have already set the IOSB. + * therefore, we can do async_set_result() directly and let the client skip + * waiting on wait_handle. */ async_set_result( &async->obj, status, information ); diff --git a/server/protocol.def b/server/protocol.def index 2e57c91ccad..5eb63db3091 100644 --- a/server/protocol.def +++ b/server/protocol.def @@ -2169,6 +2169,7 @@ enum message_type obj_handle_t handle; /* wait handle */ apc_param_t information; /* IO_STATUS_BLOCK Information */ unsigned int status; /* completion status */ + int mark_pending; /* set async to pending before completion? */ @REPLY obj_handle_t handle; /* wait handle, or NULL if closed */ @END -- 2.34.1

1 0