wine-devel

wine-devel@winehq.org

2 participants
84469 discussions

Re: lpOverlapped = NULL
by Mike McCormack 09 Jan '02

09 Jan '02

Hi Martin, Sure, feel free to change it. I used the lpOverlapped to identify the async operation, but any pointer will do. The only reason for not using async_private* was to have some way to verify that the pointer was valid... Mike Original message from: Martin Wilck <Martin.Wilck(a)fujitsu-siemens.com> > > >Mike, > >I should have been aware of this before :-/ > >As mentioned in my other mail, >** Winsock allows NULL pointers for lpOverlapped. ** > >AFAICS, Mike, your server-side code uses the overlapped field >only for identifying a certain request. This will obviously >break if the overlapped field may be NULL. > >Question: Would it hurt to use the async_private pointer of a >request rather than the overlapped pointer to pass to the >server? If no, I guess that would be the way to go, because >the async_private structs are allocated by us and therefore necessarily >different and non-NULL. > >Wrt to the status and "bytes read", I already have a pretty elegant >solution which keeps the overlapped struct outside the generic code >(scheduler and server). > >Please comment, >Martin ------------------------------------------ mailto:Mike_McCormack@start.com.au ph +82 16 430 0425 __________________________________________________________________ Get your free Australian email account at http://www.Looksmart.com.au

1 0

lpOverlapped = NULL
by Martin Wilck 09 Jan '02

09 Jan '02

Mike, I should have been aware of this before :-/ As mentioned in my other mail, ** Winsock allows NULL pointers for lpOverlapped. ** AFAICS, Mike, your server-side code uses the overlapped field only for identifying a certain request. This will obviously break if the overlapped field may be NULL. Question: Would it hurt to use the async_private pointer of a request rather than the overlapped pointer to pass to the server? If no, I guess that would be the way to go, because the async_private structs are allocated by us and therefore necessarily different and non-NULL. Wrt to the status and "bytes read", I already have a pretty elegant solution which keeps the overlapped struct outside the generic code (scheduler and server). Please comment, Martin -- Martin Wilck Phone: +49 5251 8 15113 Fujitsu Siemens Computers Fax: +49 5251 8 20409 Heinz-Nixdorf-Ring 1 mailto:Martin.Wilck@Fujitsu-Siemens.com D-33106 Paderborn http://www.fujitsu-siemens.com/primergy

1 0

struct async_private needs generalization for Winsock
by Martin Wilck 09 Jan '02

09 Jan '02

Hi Mike, all, as I pointed out before, we must change the async_private type such that it accomodates the different needs of normal (e.g. ReadFileEx()) and Winsock (e.g. WSARecv()) async requests. The "generic" functions in scheduler/synchro.c must be able to handle these structures without knowing what type of async_private structure they have been passed. I try to summarize the differences between the two request types here: 1. Winsock async requests have a WSAOVERLAPPED struct rather than OVERLAPPED. Unlike File I/O, the routines complete if any positive number of bytes is read (there's no such thing as BytesToRead). 2. Unlike File I/O, The WSAOVERLAPPED struct is *not* accessed in any way, except for the hEvent field. The other fields do not carry useful information and are, according to the Winsock2 specs, reserved to service providers, i.e. we are not allowed to write to them. 3. Winsock functions take a scatter/gather array of type WSABUF* instead of a single buffer, and a number of WSABUFs instead of the number of bytes to read. 4. The completion routine takes an additional DWORD argument. 5. In the case of WSARecvFrom()/WSASendTo(), the socket address and length parameters passed by the user must be stored and filled in at completion. [These are the differences I've come across so far, there may be more.] I'd therefore recommend the following, similar to the treatment of "struct object" and derived types in the server code: struct async_private; typedef void (*async_handler)(struct async_private *ovp); struct async_private { unsigned int type; /* file/socket, read/write/... */ HANDLE handle; int fd; struct async_private *prev; struct async_private *next; async_handler func; } struct async_fileio { struct async_private async; char *buffer; int count; LPOVERLAPPED lpOverlapped; LPOVERLAPPED_COMPLETION_ROUTINE completion_func; } struct async_winsock { struct async_private async; LPWSABUF wsabuf; int n_wsabufs; LPWSAOVERLAPPED overlapped; LPWSAOVERLAPPED_COMPLETION_ROUTINE completion_func; struct WS_sockaddr* addr; /* User-specified pointers */ int *addrlen; int flags; /* Return values */ int NumberOfBytesRead; int status; /* cannot be stored in overlapped */ } To my knowledge, there are only these two types of Asynchronous requests. Thus, using a "struct async_ops" like "struct object_ops" in the server code would probably be too much; the type field could instead be used for distinction between the two request types and the "generic" code in scheduler/synchro.c could incorporate respective case distinctions. To my experience this runs faster than code with function pointers if we have only two cases. Actually, we might even do without the "async_handler" field and decide which function to call only according to the "type" field, but that's a cosmetic issue. Unless anybody objects, I'll soon submit a patch incorporating this approach. Comments welcome, Martin -- Martin Wilck Phone: +49 5251 8 15113 Fujitsu Siemens Computers Fax: +49 5251 8 20409 Heinz-Nixdorf-Ring 1 mailto:Martin.Wilck@Fujitsu-Siemens.com D-33106 Paderborn http://www.fujitsu-siemens.com/primergy

4 7

Re: We *really* need a development model change !
by Gerard Patel 09 Jan '02

09 Jan '02

At 04:44 PM 08/01/2002 -0500, you wrote: <snip> >(Another Pythoner, cool :) ) This has given me an idea - while I don't expect it to be used in Wine, I will try to write my own test progs with it : use the *windows* python interpreter under Wine. >From the doc, it's possible to call any win32 api from it using a 'calldll' interface. If it works well, I won't need a 'test framework' - just use the standard, available tool. Gerard

4 4

Immediate IO in ReadFile / WriteFile
by Martin Wilck 09 Jan '02

09 Jan '02

Hi Mike, I just had the following thought on the immediate IO in ReadFile() / WriteFile() that you implemented: Consider the following situation: 1. There is no data available for reading on some fd (serial port, pipe, socket). 2. An application calls ReadFile() in overlapped mode. since no data is there, an overlapped request is scheduled. 3. Data becomes available before the app issues a wait request of any kind. 4. The application starts another ReadFile() request. Now data is there, and will be read immediately. No async request is scheduled. 5. Eventually, more data will become available, and the async request first scheduled will also return. However, this basically destroys the order of the requests. I am unsure how Windows would behave in such a case, and it is certainly difficult to provide a test case for this situation. It can occur, though, and my guess is this is the wrong behaviour. Thus, probably before doing the immediate read it should be checked if there are asynchronous requests already scheduled for reading, and if yes, ReadFile() should _not_ try to read immediately but go pending. Of course, this check would require a server call. The same reasoning applies for writing, of course. Things become even more confusing if you think of several threads reading from the same file. A very simple workaround would be to go pending unconditonally for overlapped requests. Although not optimal, this would be certain not to corrupt data. What do you think? Martin -- Martin Wilck Phone: +49 5251 8 15113 Fujitsu Siemens Computers Fax: +49 5251 8 20409 Heinz-Nixdorf-Ring 1 mailto:Martin.Wilck@Fujitsu-Siemens.com D-33106 Paderborn http://www.fujitsu-siemens.com/primergy

1 1

Re: wine/ files/file.c server/file.c
by Martin Wilck 09 Jan '02

09 Jan '02

On 8 Jan 2002, Alexandre Julliard wrote: > I'm not sure 002 is a good idea. It seems to encourage doing > distinctions between file types where we should instead try to handle > them in a common way. The whole point of the FD_TYPE_XYZ distinction, as it currently is, is that at some places (actually, only in WriteFile() and ReadFile()), a distinction according to fd type _must_ be made. I am not sure if Windows allows overlapped IO on the console, for example, but if yes, this patch would be a prerequisite to implement that. I just wanted to make this distinction more logical. I think it is necessary to separate fd types and flags. The distinction between regular files, pipes, sockets, ... that the patch introduces is currently not needed, but doesn't hurt IMO. If you wish, I can strip the patch down to its essential part (have separate flags for overlapped and timeout flags). I guess I could also proceed entirely without this patch, if you seriously reject it as a whole, but I fear this will bring trouble sooner or later. Martin -- Martin Wilck Phone: +49 5251 8 15113 Fujitsu Siemens Computers Fax: +49 5251 8 20409 Heinz-Nixdorf-Ring 1 mailto:Martin.Wilck@Fujitsu-Siemens.com D-33106 Paderborn http://www.fujitsu-siemens.com/primergy

2 2

Re: We *really* need a development model change !
by Andriy Palamarchuk 09 Jan '02

09 Jan '02

Spent a couple of days trying to port my test to C unit testing framework "Check". Simple tests work fine. Problems start when I try to use multithreaded Wine tests. In one case the framework stucks on reading from pipe. Other code layouts have other issues. The cause may be in incorrect pipes handling, multi-process tests handling or interaction with Wine thread implementation. I remember I also had some problems with Perl multithreaded tests... I attached my unit test example, almost identical to the one I published before for Perl framework. There is also a whole working directory. If you want to play with the it, just unpack it to wine/programs directory. Andriy Palamarchuk __________________________________________________ Do You Yahoo!? Send FREE video emails in Yahoo! Mail! http://promo.yahoo.com/videomail/

5 14

Re: Listview ctrl
by Gerard Patel 09 Jan '02

09 Jan '02

At 09:31 AM 08/01/2002 -0500, you wrote: <snip> >So, after much work and no testing, here is a first try at doing that. The >good news is that it seems it compiles. The bad news is that I have no >idea how to test it. So any suggestions on how to test the listview >control, or any help with the testing is greatly appreciated, as well as >any comments on the patch itself. > I have tested it with some of my test apps and I see a few regressions. With rowlist.exe that comes with Vc++, using 'large icons' and 'full details' only the first character of the text is displayed. The little program found at www.timex.com (IIRC it installs correctly under Wine) displays another problem (maybe related) : columns other than the first can be not displayed in some cases. Gerard

1 0

Wine & non X86 processors
by Hetz Ben Hamo 09 Jan '02

09 Jan '02

Hi, A friend of mine who uses sparc machine wants to write a pentium emulator & use wine with it to run some apps... His question is - are there any special notes about wine in big endian? does he need to make lots of changes? if so, in whats parts in general? Thanks, Hetz

3 2

Linux Today - Tonight Live: New Directions for a New Year. We talk about WINE >> on The Linux Show
by Jeremy White 08 Jan '02

08 Jan '02

FYI, they're apparently going to be talking about Wine tonight. Jer <http://linuxtoday.com/news_story.php3?ltsn=2002-01-08-024-20-OP-CY>

1 0

RFC: NamedPipes emulation API
by Luke Kenneth Casson Leighton 08 Jan '02

08 Jan '02

this is an RFC document describing an API that may be used to provide Windows NT Style "NamedPipe" functionality. it will be of interest to samba, wine, freedce and win32 developers wishing to port applications to unix (via the wine win32 implementation of NamedPipes). the functionality described in this document is equivalent to, if you have access to the MSDN, the following (incomplete list of) Win32 functions: - CreateNamedPipe (returns a file handle) - TransactNamedPipe - a read operation on the named pipe file handle, above - a write operation on the named pipe file handle, above - a close operation on the named pipe file handle, above - WaitNamedPipe there follows a list of function declarations. these are samba-specific functions that are only of use to SAMBA 3 developers. int named_pipe_create( connection_struct *connection, /* contains all info needed, e.g. user security context */ const char *pipe_name, /* e.g. \\server\PIPE\pipe_name or \\.\PIPE\pipe_name */ ); this function returns -1 on failure or an SMB file index (fnum) on success. it is called when an SMBopenX or NTcreateX is received. from memory, the access point is api_pipe_open(), in pipe.c or reply.c. STATUS named_pipe_transact(connection_struct *connection, int fnum, /* the SMB file index from the named_pipe_create */ const char *transact_data_to_send, /* smb data received, to send to NamedPipe service */ int transact_data_length, /* zero or more bytes */ char *transact_receive_buffer, /* NamedPipe data received, to be sent over SMB */ int *transact_receive_length); /* zero or more bytes */ this function returns an NT status code. if there is more data to be received, then the function must return ERROR_MORE_DATA. NT_STATUS_OK is returned on success. from memory, the access point is api_pipe_readwrite() or reply_trans2(), or maybe reply_pipe_trans(). i don't recall which. again, it'll be in pipe.c or reply.c. STATUS named_pipe_read(connection_struct *connection, int fnum, /* the SMB file index from the named_pipe_create */ char *transact_receive_buffer, /* NamedPipe data received, to be sent over SMB */ int *transact_receive_length); /* zero or more bytes */ this function returns an NT status code. NT_STATUS_OK is returned on success. this one's to be called by pipe_read() (api_pipe_read()?) STATUS named_pipe_write(connection_struct *connection, int fnum, /* the SMB file index from the named_pipe_create */ const char *transact_data_to_send, /* smb data received, to send to NamedPipe service */ int transact_data_length); /* zero or more bytes */ this function returns an NT status code. NT_STATUS_OK is returned on success. this one's to be called by pipe_write() (api_pipe_write()?) the use of this API is quite simple, and provides a clean "cut-off" point between samba and all NamedPipe transactions. lkcl

1 0

RFC: new async_private implementation
by Martin Wilck 08 Jan '02

08 Jan '02

The patch below implements my ideas from the "struct async_private needs generalization for Winsock" thread. It's a patch against CVS 2002-01-08. The Winsock stuff is still missing, that's the next step. This is only a trivial recoding of the File IO (compiles, successfully tested with my async file IO test code), whith the "generic" fields in async_private and the "fileio-specific" fields in the rest of the async_fileio structure. Note: Since it seems agreeable to use the lpOverlapped in Winsock the same way as for file IO, I made lpOverlapped a "generic" field. Mike (or other people interested), if you can test this against your serial port code I'd be grateful. Martin Patch file: async_private.diff Purpose: New winsock-friendly async_private implementation Created: 2002-01-08 Applies: CVS 2002-01-08 ChangeLog: wine/include/file.h: struct async_private: - remove buffer, count, completion_func fields. - add call_completion field (callback function that calls the completion function). struct async_fileio: - "Derived" from async_private, contains. remove buffer, count, completion_func fields. wine/scheduler/synchro.c: call_completion_routine(): function removed. finish_async(): use ovp->completion_func instead of call_completion_routine. check_async_list(): Initialize ovp. wine/files/file.c: fileio_call_completion_func(): New function, FileIO-specific callback. FILE_AsyncReadService(): - use struct async_fileio for specific fields. FILE_AsyncWriteService(): - dito. FILE_ReadFileEx(): - use async_fileio struct for overlapped initialization. - initialize ovp->call_completion field. FILE_WriteFileEx(): - dito. dlls/kernel/comm.c: COMM_WaitCommEventService(): - use struct async_fileio for "buffer" field. COMM_WaitCommEvent(): - see FILE_ReadFileEx(). Martin Wilck <Martin.Wilck(a)Fujitsu-Siemens.com> diff -ruX diffignore CVS/wine/dlls/kernel/comm.c MW/wine/dlls/kernel/comm.c --- CVS/wine/dlls/kernel/comm.c Tue Jan 8 17:43:25 2002 +++ MW/wine/dlls/kernel/comm.c Tue Jan 8 17:47:41 2002 @@ -1515,11 +1515,12 @@ static void COMM_WaitCommEventService(async_private *ovp) { LPOVERLAPPED lpOverlapped = ovp->lpOverlapped; + async_fileio *fileio = (async_fileio*) ovp; TRACE("overlapped %p\n",lpOverlapped); /* FIXME: detect other events */ - *ovp->buffer = EV_RXCHAR; + *fileio->buffer = EV_RXCHAR; lpOverlapped->Internal = STATUS_SUCCESS; } @@ -1536,7 +1537,7 @@ LPOVERLAPPED lpOverlapped) /* [in/out] for Asynchronous waiting */ { int fd,ret; - async_private *ovp; + async_fileio *ovp; if(!lpOverlapped) { @@ -1556,26 +1557,27 @@ if(fd<0) return FALSE; - ovp = (async_private *) HeapAlloc(GetProcessHeap(), 0, sizeof (async_private)); + ovp = (async_fileio *) HeapAlloc(GetProcessHeap(), 0, sizeof (async_fileio)); if(!ovp) { close(fd); return FALSE; } - ovp->lpOverlapped = lpOverlapped; - ovp->func = COMM_WaitCommEventService; + ovp->async.handle = hFile; + ovp->async.fd = fd; + ovp->async.type = ASYNC_TYPE_WAIT; + ovp->async.lpOverlapped = lpOverlapped; + ovp->async.func = COMM_WaitCommEventService; + ovp->async.call_completion = NULL; ovp->buffer = (char *)lpdwEvents; - ovp->fd = fd; ovp->count = 0; ovp->completion_func = 0; - ovp->type = ASYNC_TYPE_WAIT; - ovp->handle = hFile; - ovp->next = NtCurrentTeb()->pending_list; - ovp->prev = NULL; - if(ovp->next) - ovp->next->prev=ovp; - NtCurrentTeb()->pending_list = ovp; + ovp->async.next = NtCurrentTeb()->pending_list; + ovp->async.prev = NULL; + if(ovp->async.next) + ovp->async.next->prev=&ovp->async; + NtCurrentTeb()->pending_list = &ovp->async; /* start an ASYNCHRONOUS WaitCommEvent */ SERVER_START_REQ( register_async ) diff -ruX diffignore CVS/wine/files/file.c MW/wine/files/file.c --- CVS/wine/files/file.c Tue Jan 8 17:43:51 2002 +++ MW/wine/files/file.c Tue Jan 8 17:47:41 2002 @@ -1298,6 +1298,21 @@ } /*********************************************************************** + * fileio_call_completion_func (INTERNAL) + */ +void CALLBACK fileio_call_completion_func (ULONG_PTR data) +{ + async_fileio *ovp = (async_fileio*) data; + TRACE ("data: %p\n", ovp); + + ovp->completion_func(ovp->async.lpOverlapped->Internal, + ovp->async.lpOverlapped->InternalHigh, + ovp->async.lpOverlapped); + ovp->completion_func=NULL; + HeapFree(GetProcessHeap(), 0, ovp); +} + +/*********************************************************************** * FILE_AsyncReadService (INTERNAL) * * This function is called while the client is waiting on the @@ -1305,18 +1320,19 @@ */ static void FILE_AsyncReadService(async_private *ovp) { + async_fileio *fileio = (async_fileio*) ovp; LPOVERLAPPED lpOverlapped = ovp->lpOverlapped; int result, r; int already = lpOverlapped->InternalHigh; - TRACE("%p %p\n", lpOverlapped, ovp->buffer ); + TRACE("%p %p\n", lpOverlapped, fileio->buffer ); /* check to see if the data is ready (non-blocking) */ - result = pread (ovp->fd, &ovp->buffer[already], ovp->count - already, + result = pread (ovp->fd, &fileio->buffer[already], fileio->count - already, OVERLAPPED_OFFSET (lpOverlapped) + already); if ((result < 0) && (errno == ESPIPE)) - result = read (ovp->fd, &ovp->buffer[already], ovp->count - already); + result = read (ovp->fd, &fileio->buffer[already], fileio->count - already); if ( (result<0) && ((errno == EAGAIN) || (errno == EINTR))) { @@ -1336,7 +1352,7 @@ lpOverlapped->InternalHigh += result; TRACE("read %d more bytes %ld/%d so far\n",result,lpOverlapped->InternalHigh,ovp->count); - if(lpOverlapped->InternalHigh < ovp->count) + if(lpOverlapped->InternalHigh < fileio->count) r = STATUS_PENDING; else r = STATUS_SUCCESS; @@ -1352,7 +1368,7 @@ LPOVERLAPPED overlapped, LPOVERLAPPED_COMPLETION_ROUTINE lpCompletionRoutine) { - async_private *ovp; + async_fileio *ovp; int fd; TRACE("file %d to buf %p num %ld %p func %p\n", @@ -1372,7 +1388,7 @@ return FALSE; } - ovp = (async_private *) HeapAlloc(GetProcessHeap(), 0, sizeof (async_private)); + ovp = (async_fileio *) HeapAlloc(GetProcessHeap(), 0, sizeof (async_fileio)); if(!ovp) { TRACE("HeapAlloc Failed\n"); @@ -1380,21 +1396,23 @@ close(fd); return FALSE; } - ovp->lpOverlapped = overlapped; - ovp->count = bytesToRead; + + ovp->async.handle = hFile; + ovp->async.fd = fd; + ovp->async.type = ASYNC_TYPE_READ; + ovp->async.lpOverlapped = overlapped; + ovp->async.func = FILE_AsyncReadService; + ovp->async.call_completion = (lpCompletionRoutine ? fileio_call_completion_func : NULL); ovp->completion_func = lpCompletionRoutine; - ovp->func = FILE_AsyncReadService; ovp->buffer = buffer; - ovp->fd = fd; - ovp->type = ASYNC_TYPE_READ; - ovp->handle = hFile; + ovp->count = bytesToRead; /* hook this overlap into the pending async operation list */ - ovp->next = NtCurrentTeb()->pending_list; - ovp->prev = NULL; - if(ovp->next) - ovp->next->prev = ovp; - NtCurrentTeb()->pending_list = ovp; + ovp->async.next = NtCurrentTeb()->pending_list; + ovp->async.prev = NULL; + if(ovp->async.next) + ovp->async.next->prev = &ovp->async; + NtCurrentTeb()->pending_list = &ovp->async; if ( !FILE_StartAsync(hFile, overlapped, ASYNC_TYPE_READ, bytesToRead, STATUS_PENDING) ) { @@ -1550,18 +1568,19 @@ */ static void FILE_AsyncWriteService(struct async_private *ovp) { + async_fileio *fileio = (async_fileio *) ovp; LPOVERLAPPED lpOverlapped = ovp->lpOverlapped; int result, r; int already = lpOverlapped->InternalHigh; - TRACE("(%p %p)\n",lpOverlapped,ovp->buffer); + TRACE("(%p %p)\n",lpOverlapped,fileio->buffer); /* write some data (non-blocking) */ - result = pwrite(ovp->fd, &ovp->buffer[already], ovp->count - already, + result = pwrite(ovp->fd, &fileio->buffer[already], fileio->count - already, OVERLAPPED_OFFSET (lpOverlapped) + already); if ((result < 0) && (errno == ESPIPE)) - result = write(ovp->fd, &ovp->buffer[already], ovp->count - already); + result = write(ovp->fd, &fileio->buffer[already], fileio->count - already); if ( (result<0) && ((errno == EAGAIN) || (errno == EINTR))) { @@ -1578,9 +1597,9 @@ lpOverlapped->InternalHigh += result; - TRACE("wrote %d more bytes %ld/%d so far\n",result,lpOverlapped->InternalHigh,ovp->count); + TRACE("wrote %d more bytes %ld/%d so far\n",result,lpOverlapped->InternalHigh,fileio->count); - if(lpOverlapped->InternalHigh < ovp->count) + if(lpOverlapped->InternalHigh < fileio->count) r = STATUS_PENDING; else r = STATUS_SUCCESS; @@ -1596,7 +1615,7 @@ LPOVERLAPPED overlapped, LPOVERLAPPED_COMPLETION_ROUTINE lpCompletionRoutine) { - async_private *ovp; + async_fileio *ovp; TRACE("file %d to buf %p num %ld %p func %p stub\n", hFile, buffer, bytesToWrite, overlapped, lpCompletionRoutine); @@ -1616,34 +1635,35 @@ return FALSE; } - ovp = (async_private*) HeapAlloc(GetProcessHeap(), 0, sizeof (async_private)); + ovp = (async_fileio*) HeapAlloc(GetProcessHeap(), 0, sizeof (async_fileio)); if(!ovp) { TRACE("HeapAlloc Failed\n"); SetLastError(ERROR_NOT_ENOUGH_MEMORY); return FALSE; } - ovp->lpOverlapped = overlapped; - ovp->func = FILE_AsyncWriteService; + ovp->async.handle = hFile; + ovp->async.fd = FILE_GetUnixHandle( hFile, GENERIC_WRITE ); + ovp->async.type = ASYNC_TYPE_WRITE; + ovp->async.lpOverlapped = overlapped; + ovp->async.func = FILE_AsyncWriteService; + ovp->async.call_completion = (lpCompletionRoutine ? fileio_call_completion_func : NULL); ovp->buffer = (LPVOID) buffer; ovp->count = bytesToWrite; ovp->completion_func = lpCompletionRoutine; - ovp->fd = FILE_GetUnixHandle( hFile, GENERIC_WRITE ); - ovp->type = ASYNC_TYPE_WRITE; - ovp->handle = hFile; - if(ovp->fd <0) + if(ovp->async.fd <0) { HeapFree(GetProcessHeap(), 0, ovp); return FALSE; } /* hook this overlap into the pending async operation list */ - ovp->next = NtCurrentTeb()->pending_list; - ovp->prev = NULL; - if(ovp->next) - ovp->next->prev = ovp; - NtCurrentTeb()->pending_list = ovp; + ovp->async.next = NtCurrentTeb()->pending_list; + ovp->async.prev = NULL; + if(ovp->async.next) + ovp->async.next->prev = &ovp->async; + NtCurrentTeb()->pending_list = &ovp->async; return TRUE; } diff -ruX diffignore CVS/wine/include/file.h MW/wine/include/file.h --- CVS/wine/include/file.h Tue Jan 8 17:43:45 2002 +++ MW/wine/include/file.h Tue Jan 8 17:47:41 2002 @@ -33,20 +33,31 @@ /* overlapped private structure */ struct async_private; -typedef void (*async_handler)(struct async_private *ovp); + +typedef void (*async_handler) (struct async_private *ovp); +typedef void CALLBACK (*async_call_completion_func) (ULONG_PTR data); + +extern void CALLBACK fileio_call_completion_func (ULONG_PTR data); + typedef struct async_private { - LPOVERLAPPED lpOverlapped; - HANDLE handle; - int fd; - char *buffer; - async_handler func; - int count; - int type; - LPOVERLAPPED_COMPLETION_ROUTINE completion_func; - struct async_private *next; - struct async_private *prev; + HANDLE handle; + int fd; + int type; + LPOVERLAPPED lpOverlapped; + async_handler func; + async_call_completion_func call_completion; + struct async_private *next; + struct async_private *prev; } async_private; + +typedef struct async_fileio +{ + async_private async; + LPOVERLAPPED_COMPLETION_ROUTINE completion_func; + char *buffer; + int count; +} async_fileio; extern void WINAPI check_async_list(LPOVERLAPPED ov, DWORD status); extern void finish_async(struct async_private *ovp, DWORD status); diff -ruX diffignore CVS/wine/scheduler/synchro.c MW/wine/scheduler/synchro.c --- CVS/wine/scheduler/synchro.c Tue Jan 8 17:43:37 2002 +++ MW/wine/scheduler/synchro.c Tue Jan 8 17:47:41 2002 @@ -36,25 +36,14 @@ } } -static void CALLBACK call_completion_routine(ULONG_PTR data) -{ - async_private* ovp = (async_private*)data; - - ovp->completion_func(ovp->lpOverlapped->Internal, - ovp->lpOverlapped->InternalHigh, - ovp->lpOverlapped); - ovp->completion_func=NULL; - HeapFree(GetProcessHeap(), 0, ovp); -} - void finish_async(async_private *ovp, DWORD status) { ovp->lpOverlapped->Internal=status; - /* call ReadFileEx/WriteFileEx's overlapped completion function */ - if(ovp->completion_func) + /* call overlapped completion function */ + if(ovp->call_completion) { - QueueUserAPC(call_completion_routine,GetCurrentThread(),(ULONG_PTR)ovp); + QueueUserAPC(ovp->call_completion,GetCurrentThread(),(ULONG_PTR)ovp); } /* remove it from the active list */ @@ -70,7 +59,7 @@ ovp->prev=NULL; close(ovp->fd); - if(!ovp->completion_func) HeapFree(GetProcessHeap(), 0, ovp); + if(!ovp->call_completion) HeapFree(GetProcessHeap(), 0, ovp); } /*********************************************************************** @@ -80,7 +69,7 @@ */ void WINAPI check_async_list(LPOVERLAPPED overlapped, DWORD status) { - async_private *ovp; + async_private *ovp = NULL; /* fprintf(stderr,"overlapped %p status %x\n",overlapped,status); */

1 0

Patch: BUG in thread_queue_apc()
by Martin Wilck 08 Jan '02

08 Jan '02

The patch below fixes a bug in server/thread.c that causes APCs to be "forgotten". The patch is against today's CVS (2002-01-08). I came across this by testing async file I/O. When submitting several async IO requests, the IO completion routine was called only once for the first submitted request. I cannot oversee if this can break anything elsewhere, but I strongly believe the behaviour with this patch is more sane than without. Martin -- Martin Wilck Phone: +49 5251 8 15113 Fujitsu Siemens Computers Fax: +49 5251 8 20409 Heinz-Nixdorf-Ring 1 mailto:Martin.Wilck@Fujitsu-Siemens.com D-33106 Paderborn http://www.fujitsu-siemens.com/primergy Index: server/thread.c =================================================================== RCS file: /home/wine/wine/server/thread.c,v retrieving revision 1.72 diff -u -r1.72 thread.c --- server/thread.c 2001/12/20 00:07:09 1.72 +++ server/thread.c 2002/01/08 13:35:17 @@ -562,6 +562,8 @@ queue->head = apc; wake_thread( thread ); } + else apc->prev->next = apc; + return 1; }

1 0

Re: struct async_private needs generalization for Winsock
by Mike McCormack 08 Jan '02

08 Jan '02

Hi Martin, That's pretty quick and thorough work! >I try to summarize the differences between the two request types here: > >1. Winsock async requests have a WSAOVERLAPPED struct rather than > OVERLAPPED. Unlike File I/O, the routines complete if any > positive number of bytes is read (there's no such thing as > BytesToRead). That case is similar to serial I/O when timeouts are (interval=-1,const=0,total=0), isn't it? The file I/O code doesn't really handle it properly yet either, as it just falls into a busy loop in the original ReadFile... hmmm. >2. Unlike File I/O, The WSAOVERLAPPED struct is *not* accessed in any Looks like you sorted this out in the other thread. i'd suggest you use the WSAOVERLAPPED fields in the same way as OVERLAPPED where ever possible, or not use them at all. i think you need the completion status in the Internal member... HasOverlappedCompleted is a macro that checks there. >[These are the differences I've come across so far, there may be more.] The way i've done the serial overlapped code so far is to implement what i *think* should be done, then wait for people to complain about their programs not working. You won't have to wait too long... >I'd therefore recommend the following, similar to the treatment of >"struct object" and derived types in the server code: > >struct async_private; >typedef void (*async_handler)(struct async_private *ovp); .. That looks OK to me. i'm not going to complain about it too much if i'm not doing the work to implement it ;-) Mike ------------------------------------------ mailto:Mike_McCormack@start.com.au ph +82 16 430 0425 __________________________________________________________________ Get your free Australian email account at http://www.Looksmart.com.au

2 1

RE: Debugger on Solaris
by Patrik Stridvall 07 Jan '02

07 Jan '02

[I'm not sure what you are doing but...] > Robert Lunnon <bobNoSpaM(a)yarrabee.net.au> writes: > > > Here is what the Sun man page says about Ptrace, with > apologies to sun, you > > can find this page at docs.sun.com. Not the comment for > call 0 (TRACEME) and > > ESRCH in the error section > > But there has to be some way to attach to a running process. Yes, ptrace(PTRACE_ATTACH, ...) should work I believe. Haven't tried though. > What > happens when you do an attach in gdb? Does it do a ptrace? I'm not sure. Here is a trace of gdb trying to attach to a bash process (pid 11664).

2 1

Re: Debugger on Solaris
by Robert Lunnon 07 Jan '02

07 Jan '02

> On Fri, 4 Jan 2002 10:49, Alexandre Julliard wrote: > > Robert Lunnon <bobNoSpaM(a)yarrabee.net.au> writes: > > > Without calling ptrace (TRACEME,...) Solaris returns ESRCH even though > > > the process exists which causes the wineserver to assume the process is > > > defunct and delete the app, apparently causing the deadlock (This > > > deadlock will have to be fixed too !). > > > > Are you sure this is not rather an issue of process id vs. thread id? > > I would be surprised if Solaris required TRACEME in this case, since > > this would mean you couldn't attach a debugger to a running process. > >it seems something is missing from your mail <grin> > >-- >Alexandre Julliard >julliard(a)winehq.com Some strange behaviour of my mail client is removing edits from replies Sorry all. Here is what the Sun man page says about Ptrace, with apologies to sun, you can find this page at docs.sun.com. Not the comment for call 0 (TRACEME) and ESRCH in the error section System Calls ptrace(2) NAME ptrace - allows a parent process to control the execution of a child process SYNOPSIS #include <unistd.h> #include <sys/types.h> int ptrace(int request, pid_t pid, int addr, int data); DESCRIPTION The ptrace() function allows a parent process to control the execution of a child process. Its primary use is for the implementation of breakpoint debugging. The child process behaves normally until it encounters a signal (see signal(3HEAD)), at which time it enters a stopped state and its parent is notified via the wait(2) function. When the child is in the stopped state, its parent can examine and modify its "core image" using ptrace(). Also, the parent can cause the child either to terminate or continue, with the possibility of ignoring the signal that caused it to stop. The request argument determines the action to be taken by ptrace() and is one of the following: 0 This request must be issued by the child process if it is to be traced by its parent. It turns on the child's trace flag that stipulates that the child should be left in a stopped state on receipt of a signal rather than the state specified by func (see signal(3C)). The pid, addr, and data arguments are ignored, and a return value is not defined for this request. Peculiar results ensue if the parent does not expect to trace the child. The remainder of the requests can only be used by the parent process. For each, pid is the process ID of the child. The child must be in a stopped state before these requests are made. 1, 2 With these requests, the word at location addr in the address space of the child is returned to the parent process. If instruction and data space are separated, request 1 returns a word from instruction space, and request 2 returns a word from data space. If instruc- tion and data space are not separated, either request 1 or request 2 may be used with equal results. The data argument is ignored. These two requests fail if addr is not the start address of a word, in which case -1 is returned to the parent process and the parent's errno is set to EIO. SunOS 5.8 Last change: 5 Jul 1990 1 System Calls ptrace(2) 3 With this request, the word at location addr in the child's user area in the system's address space (see <sys/user.h>) is returned to the parent process. The data argument is ignored. This request fails if addr is not the start address of a word or is outside the user area, in which case -1 is returned to the parent process and the parent's errno is set to EIO. 4, 5 With these requests, the value given by the data argu- ment is written into the address space of the child at location addr. If instruction and data space are separated, request 4 writes a word into instruction space, and request 5 writes a word into data space. If instruction and data space are not separated, either request 4 or request 5 may be used with equal results. On success, the value written into the address space of the child is returned to the parent. These two requests fail if addr is not the start address of a word. On failure -1 is returned to the parent process and the parent's errno is set to EIO. 6 With this request, a few entries in the child's user area can be written. data gives the value that is to be written and addr is the location of the entry. The few entries that can be written are the general regis- ters and the condition codes of the Processor Status Word. 7 This request causes the child to resume execution. If the data argument is 0, all pending signals including the one that caused the child to stop are canceled before it resumes execution. If the data argument is a valid signal number, the child resumes execution as if it had incurred that signal, and any other pending signals are canceled. The addr argument must be equal to 1 for this request. On success, the value of data is returned to the parent. This request fails if data is not 0 or a valid signal number, in which case -1 is returned to the parent process and the parent's errno is set to EIO. 8 This request causes the child to terminate with the same consequences as exit(2). 9 This request sets the trace bit in the Processor Status Word of the child and then executes the same steps as listed above for request 7. The trace bit causes an interrupt on completion of one machine instruction. This effectively allows single stepping of the child. SunOS 5.8 Last change: 5 Jul 1990 2 System Calls ptrace(2) To forestall possible fraud, ptrace() inhibits the set- user-ID facility on subsequent calls to one of the exec fam- ily of functions (see exec(2)). If a traced process calls one of the exec functions, it stops before executing the first instruction of the new image showing signal SIGTRAP. ERRORS The ptrace() function will fail if: EIO The request argument is an illegal number. EPERM The effective user of the calling process is not super-user. ESRCH The pid argument identifies a child that does not exist or has not executed a ptrace() call with request 0. SEE ALSO exec(2), exit(2), wait(2), signal(3C), signal(3HEAD)

2 1

Re: cups printing quicken
by Marcus Meissner 07 Jan '02

07 Jan '02

On Sun, Jan 06, 2002 at 08:06:33AM +0000, Alex Newman wrote: > Hello wine people, this is my first posting here. I am using the gimp > printing ppd,cups and lpr on a hpdj842c. Wine release 20011226. On Debian > woody with 2.4.17. Everything rocks so far except whenever i try to print > in quicken either 2002 or version 6 which seem to printing with the same > program. wine freaks out and dies. > > wine --debugmsg +winspool .wine/fake_windows/Program\ Files/QUICKENW/QW.EXE > was sent to > http://www.wuli.nu/users/dolemite/winedebugmsg > my .wine/config is at > http://www.wuli.nu/users/dolemite/wineconfig > > I have printed with other progs in wine and normally and it works fine. Difficult. Try running it within winedbg, so it triggers at the time the exception appears and we can see the backtrace. (Just replace 'wine' by 'winedbg' in the commandline, type 'cont' to start the program.) Ciao, Marcus

1 0

Changes regarding wnaspi32 layer ?
by ma.hoff＠t-online.de 07 Jan '02

07 Jan '02

Hi, i have some problems with the winaspi implementation (scsi cd-writer) of the current version of Wine (tried cvs one or two weeks ago ...) With the version 20010731 which is included in my SuSE 7.3 distribution it works fine (except for some graphic glitches in the application - but thats another story). I try to run Aheads Nero Burning Rom 5.0 , which refuses to detect the cd-writer with the newer versions of wine ! I heard of someone running a cvs snapshot from october 2001 which is working, too - but all later version of wine refuse to work with nero ! The error message is the following: fixme:win32:DEVICE_Open Unknown/unsupported VxD nerocd95. Try --winver nt40 or win31 ! fixme:win32:DEVICE_Open Unknown/unsupported VxD BsUDF.VXD. Try --winver nt40 or win31 ! Guess what: --winver nt40 / win31 didn't change anything :-) So it seems that its not implemented, but with older versions it worked ! Is this a known bug - or do i have to change something else when using the newer version of wine ? Or is it just broken ? Thanks in advance for any help ! Martin

2 1

implement Cancelio (updated), WriteFile update, comm16 update
by l_whitney＠juno.com 07 Jan '02

07 Jan '02

Good day! If you get this, these three patches fix a regression in 16 bit serial comms as practiced by Juno 1.49 that was intrduced by: Changes by: julliard@wine2. 01/12/31 20:11:33 Modified files: files : file.c scheduler : synchro.c Log message: Mike McCormack <mike_mccormack(a)start.com.au>R Set an overlapped's hEvent for ReadFile/WriteFile in an overlapped completion function, not in ReadFileEx or WriteFileEx. Patch: http://cvs.winehq.com/patch.py?id=1009847493736738936509937 So I'm in favor of committing them. Regards, Lawson

1 0

Patch: 005-recv.diff
by Martin Wilck 07 Jan '02

07 Jan '02

#5 in series Patch file: 005-recv.diff Purpose: Move recv() functionality to WS2_32 code Created: 2002-01-07 Applies: CVS 2002-01-07 (with 004-socket.diff applied) This patch moves the recv() functionality to WS2_32 functions that provide a superset of the functionality of their WS counterparts. For the recv() family of functions, WSARecvFrom() becomes the "work horse". All other functions (WSARecv(), recv(), recvfrom()) are only wrappers around WSARecvFrom(). To provide the scatter/gather functionality of WSARecv() / WSARecvFrom(), the (only) unix function called is recvmsg(). This means some overhead for simple recv() calls, but allows for a single general "work" code. dlls/winsock/socket.c: WS_recv(): Is now a wrapper around WSARecv(). WS_recvfrom(): Is now a wrapper around WSARecvFrom(). WSARecv(): Is now a wrapper around WSARecvFrom(). WSARecvFrom(): Holds all "real" recv() functionality (formerly in WS_recv() / WS_recvfrom()). Calls recvmsg() to provide for scatter/gather functionality. dlls/winsock/ws2_32.spec: Add stdcall entry for WSARecv(). Martin Wilck <Martin.Wilck(a)Fujitsu-Siemens.com> diff -ru -X diffignore CVS/wine/dlls/winsock/socket.c MW/wine/dlls/winsock/socket.c --- CVS/wine/dlls/winsock/socket.c Mon Jan 7 12:55:24 2002 +++ MW/wine/dlls/winsock/socket.c Mon Jan 7 12:55:37 2002 @@ -1815,34 +1815,16 @@ */ int WINAPI WS_recv(SOCKET s, char *buf, int len, int flags) { - int fd = _get_sock_fd(s); + DWORD n; + WSABUF wsabuf = { len, buf }; + DWORD dwFlags = flags; TRACE("socket %04x, buf %8x, len %d, flags %d\n", s, (unsigned)buf, len, flags); - if (fd != -1) - { - INT length; - - if (_is_blocking(s)) - { - /* block here */ - /* FIXME: OOB and exceptfds? */ - do_block(fd, 1); - } - if ((length = recv(fd, buf, len, flags)) >= 0) - { - TRACE(" -> %i bytes\n", length); - - close(fd); - _enable_event(s, FD_READ, 0, 0); - return length; - } - SetLastError(wsaErrno()); - close(fd); - } - else SetLastError(WSAENOTSOCK); - WARN(" -> ERROR\n"); - return SOCKET_ERROR; + if ( WSARecv (s, &wsabuf, 1, &n, &dwFlags, NULL, NULL) == SOCKET_ERROR ) + return SOCKET_ERROR; + else + return n; } /*********************************************************************** @@ -1860,60 +1842,14 @@ int WINAPI WS_recvfrom(SOCKET s, char *buf, INT len, int flags, struct WS_sockaddr *from, int *fromlen) { - int fd = _get_sock_fd(s); - int res; - - TRACE("socket %04x, ptr %08x, len %d, flags %d\n", s, (unsigned)buf, len, flags); -#if DEBUG_SOCKADDR - if (from) - dump_sockaddr(from); - else - DPRINTF("from = NULL\n"); -#endif - - res=SOCKET_ERROR; - if (fd != -1) - { - struct sockaddr* uaddr; - int uaddrlen; - int length; + DWORD n; + WSABUF wsabuf = { len, buf }; + DWORD dwFlags = flags; - if (_is_blocking(s)) - { - /* block here */ - /* FIXME: OOB and exceptfds */ - do_block(fd, 1); - } - - uaddr=ws_sockaddr_alloc(from,fromlen,&uaddrlen); - length=recvfrom(fd, buf, len, flags, uaddr, &uaddrlen); - if (length < 0) - { - SetLastError(wsaErrno()); - WARN(" -> ERROR\n"); - } - else if (ws_sockaddr_u2ws(uaddr,uaddrlen,from,fromlen) != 0) - { - /* The from buffer was too small, but we read the data - * anyway. Is that really bad? - */ - SetLastError(WSAEFAULT); - WARN(" -> WSAEFAULT\n"); - } - else - { - TRACE(" -> %i bytes\n", length); - _enable_event(s, FD_READ, 0, 0); - res=length; - } - close(fd); - } + if ( WSARecvFrom (s, &wsabuf, 1, &n, &dwFlags, from, fromlen, NULL, NULL) == SOCKET_ERROR ) + return SOCKET_ERROR; else - { - SetLastError(WSAENOTSOCK); - WARN(" -> WSAENOTSOCK\n"); - } - return res; + return n; } /*********************************************************************** @@ -3404,34 +3340,132 @@ /*********************************************************************** + * WSARecv (WS2_32.67) + */ +int WINAPI WSARecv (SOCKET s, LPWSABUF lpBuffers, DWORD dwBufferCount, + LPDWORD NumberOfBytesReceived, LPDWORD lpFlags, + LPWSAOVERLAPPED lpOverlapped, + LPWSAOVERLAPPED_COMPLETION_ROUTINE lpCompletionRoutine) +{ + + return WSARecvFrom (s, lpBuffers, dwBufferCount, NumberOfBytesReceived, lpFlags, + NULL, NULL, lpOverlapped, lpCompletionRoutine); + +} + +/*********************************************************************** * WSARecvFrom (WS2_32.69) */ INT WINAPI WSARecvFrom( SOCKET s, LPWSABUF lpBuffers, DWORD dwBufferCount, LPDWORD lpNumberOfBytesRecvd, LPDWORD lpFlags, struct WS_sockaddr *lpFrom, LPINT lpFromlen, LPWSAOVERLAPPED lpOverlapped, LPWSAOVERLAPPED_COMPLETION_ROUTINE lpCompletionRoutine ) + { - DWORD dwCount; - INT rc; + /* Uses recvmsg() in order to provide scatter-gather I/O */ + + struct iovec* iovec; + struct msghdr msghdr; + int fd, i, length, err = WSAENOTSOCK; + + TRACE("socket %04x, wsabuf %p, nbufs %ld, flags %ld, from %p, fromlen %ld, ovl %p, func %p\n", + s, lpBuffers, dwBufferCount, *lpFlags, lpFrom, + (lpFromlen ? *lpFromlen : -1L), + lpOverlapped, lpCompletionRoutine); - FIXME( "(%i,%p,%lu,%p,%p,%p,%p,%p,%p: stub\n", - s, lpBuffers, dwBufferCount, lpNumberOfBytesRecvd, lpFlags, - lpFrom, lpFromlen, lpOverlapped, lpCompletionRoutine ); + fd = _get_sock_fd(s); - for( dwCount = 0, rc = 0; dwCount < dwBufferCount; dwCount++ ) - { + if (fd == -1) + { + err = WSAENOTSOCK; + goto error; + } - if( ( rc = WS_recvfrom(s, lpBuffers[ dwCount ].buf, (INT)lpBuffers[ dwCount ].len, - (INT)*lpFlags, lpFrom, lpFromlen ) ) != 0 ) + /* FIXME: should this be HeapAlloc() or WS_ALLOC ? */ + iovec = WS_ALLOC ( dwBufferCount * sizeof (struct iovec) ); + if ( !iovec ) { - break; + err = WSAENOBUFS; + goto err_close; } - } + for (i = 0; i < dwBufferCount; i++) + { + iovec[i].iov_base = lpBuffers[i].buf; + iovec[i].iov_len = lpBuffers[i].len; + } - return rc; -} + msghdr.msg_name = NULL; + if ( lpFrom ) + { +#if DEBUG_SOCKADDR + dump_sockaddr (lpFrom); +#endif + + msghdr.msg_namelen = *lpFromlen; + msghdr.msg_name = ws_sockaddr_alloc (lpFrom, lpFromlen, &msghdr.msg_namelen); + } + else + msghdr.msg_namelen = 0; + + msghdr.msg_iov = iovec; + msghdr.msg_iovlen = dwBufferCount; + msghdr.msg_control = NULL; + msghdr.msg_controllen = 0; + msghdr.msg_flags = 0; + + /* FIXME: Treat overlapped IO here */ + + if (_is_blocking(s)) + { + /* block here */ + /* FIXME: OOB and exceptfds? */ + do_block(fd, 1); + } + + /* FIXME: can we support MSG_PARTIAL ? + How does it relate to recvmsg()'s msg_flags ? */ + + if ((length = recvmsg (fd, &msghdr, *lpFlags)) == -1) + { + err = wsaErrno(); + goto err_free; + } + + TRACE(" -> %i bytes\n", length); + + if ( lpFrom && ws_sockaddr_u2ws (msghdr.msg_name, msghdr.msg_namelen, lpFrom, lpFromlen) != 0 ) + { + /* The from buffer was too small, but we read the data + * anyway. Is that really bad? + */ + SetLastError ( WSAEFAULT ); + WARN ( " -> Address buffer too small\n" ); + } + + + *lpNumberOfBytesRecvd = length; + + WS_FREE (iovec); + ws_sockaddr_free ( msghdr.msg_name, lpFrom ); + close(fd); + _enable_event(s, FD_READ, 0, 0); + + return 0; + +err_free: + WS_FREE (iovec); + ws_sockaddr_free ( msghdr.msg_name, lpFrom ); + +err_close: + close (fd); + +error: + WARN(" -> ERROR %d\n", err); + SetLastError ( err ); + return SOCKET_ERROR; +} /*********************************************************************** * WSCInstallProvider (WS2_32.88) diff -ru -X diffignore CVS/wine/dlls/winsock/ws2_32.spec MW/wine/dlls/winsock/ws2_32.spec --- CVS/wine/dlls/winsock/ws2_32.spec Mon Jan 7 12:42:08 2002 +++ MW/wine/dlls/winsock/ws2_32.spec Mon Jan 7 12:55:37 2002 @@ -79,7 +79,7 @@ 64 stub WSANtohl 65 stub WSANtohs 66 stub WSAProviderConfigChange -67 stub WSARecv +67 stdcall WSARecv(long ptr long ptr ptr ptr ptr) WSARecv 68 stub WSARecvDisconnect 69 stdcall WSARecvFrom(long ptr long ptr ptr ptr ptr ptr ptr ) WSARecvFrom 70 stub WSARemoveServiceClass

1 0

Patch: 004-socket.diff
by Martin Wilck 07 Jan '02

07 Jan '02

#4 in series (incremental) Patch file: 004-socket.diff Purpose: Move socket() functionality to WS2_32 code Created: 2002-01-07 Applies: CVS 2002-01-07 (with 003-sock_basic.diff applied) This patch moves the functionality of WS_socket() to WSASocketA(). WSASocketA() accounts for WSA_FLAG_OVERLAPPED. dlls/winsock/socket.c: WS_socket(): Is now a wrapper around WSASocketA(). WSA_FLAG_OVERLAPPED is set according to the SO_OPENTYPE value. WSASocketA(): Has the functionality formerly in WS_socket(). Honours WSA_FLAG_OVERLAPPED. Martin Wilck <Martin.Wilck(a)Fujitsu-Siemens.com> diff -ru -X diffignore CVS/wine/dlls/winsock/socket.c MW/wine/dlls/winsock/socket.c --- CVS/wine/dlls/winsock/socket.c Mon Jan 7 12:45:56 2002 +++ MW/wine/dlls/winsock/socket.c Mon Jan 7 12:51:27 2002 @@ -2329,81 +2329,14 @@ */ SOCKET WINAPI WS_socket(int af, int type, int protocol) { - SOCKET ret; - TRACE("af=%d type=%d protocol=%d\n", af, type, protocol); - /* check the socket family */ - switch(af) - { -#ifdef HAVE_IPX - case WS_AF_IPX: af = AF_IPX; -#endif - case AF_INET: - case AF_UNSPEC: break; - default: SetLastError(WSAEAFNOSUPPORT); - return INVALID_SOCKET; - } - - /* check the socket type */ - switch(type) - { - case WS_SOCK_STREAM: - type=SOCK_STREAM; - break; - case WS_SOCK_DGRAM: - type=SOCK_DGRAM; - break; - case WS_SOCK_RAW: - type=SOCK_RAW; - break; - default: SetLastError(WSAESOCKTNOSUPPORT); - return INVALID_SOCKET; - } - - /* check the protocol type */ - if ( protocol < 0 ) /* don't support negative values */ - { SetLastError(WSAEPROTONOSUPPORT); return INVALID_SOCKET; } - - if ( af == AF_UNSPEC) /* did they not specify the address family? */ - switch(protocol) - { - case IPPROTO_TCP: - if (type == SOCK_STREAM) { af = AF_INET; break; } - case IPPROTO_UDP: - if (type == SOCK_DGRAM) { af = AF_INET; break; } - default: SetLastError(WSAEPROTOTYPE); return INVALID_SOCKET; - } - - SERVER_START_REQ( create_socket ) - { - req->family = af; - req->type = type; - req->protocol = protocol; - req->access = GENERIC_READ|GENERIC_WRITE|SYNCHRONIZE; - req->flags = 0U; - req->inherit = TRUE; - set_error( wine_server_call( req ) ); - ret = (SOCKET)reply->handle; - } - SERVER_END_REQ; - if (ret) - { - TRACE("\tcreated %04x\n", ret ); - return ret; - } - - if (GetLastError() == WSAEACCES) /* raw socket denied */ - { - if (type == SOCK_RAW) - MESSAGE("WARNING: Trying to create a socket of type SOCK_RAW, will fail unless running as root\n"); - else - MESSAGE("WS_SOCKET: not enough privileges to create socket, try running as root\n"); - SetLastError(WSAESOCKTNOSUPPORT); - } + /* The Winsock2 specification states that socket() always opens sockets + in overlapped mode. + FIXME: is the SO_OPENTYPE behaviour correct? */ - WARN("\t\tfailed!\n"); - return INVALID_SOCKET; + return WSASocketA ( af, type, protocol, NULL, 0, + (opentype ? 0 : WSA_FLAG_OVERLAPPED) ); } /*********************************************************************** @@ -2974,17 +2907,88 @@ LPWSAPROTOCOL_INFOA lpProtocolInfo, GROUP g, DWORD dwFlags) { + SOCKET ret; + /* FIXME: The "advanced" parameters of WSASocketA (lpProtocolInfo, - g, dwFlags) are ignored. + g, dwFlags except WSA_FLAG_OVERLAPPED) are ignored. */ TRACE("af=%d type=%d protocol=%d protocol_info=%p group=%d flags=0x%lx\n", af, type, protocol, lpProtocolInfo, g, dwFlags ); - return ( WS_socket (af, type, protocol) ); -} + /* check the socket family */ + switch(af) + { +#ifdef HAVE_IPX + case WS_AF_IPX: af = AF_IPX; +#endif + case AF_INET: + case AF_UNSPEC: break; + default: SetLastError(WSAEAFNOSUPPORT); + return INVALID_SOCKET; + } + + /* check the socket type */ + switch(type) + { + case WS_SOCK_STREAM: + type=SOCK_STREAM; + break; + case WS_SOCK_DGRAM: + type=SOCK_DGRAM; + break; + case WS_SOCK_RAW: + type=SOCK_RAW; + break; + default: SetLastError(WSAESOCKTNOSUPPORT); + return INVALID_SOCKET; + } + + /* check the protocol type */ + if ( protocol < 0 ) /* don't support negative values */ + { SetLastError(WSAEPROTONOSUPPORT); return INVALID_SOCKET; } + + if ( af == AF_UNSPEC) /* did they not specify the address family? */ + switch(protocol) + { + case IPPROTO_TCP: + if (type == SOCK_STREAM) { af = AF_INET; break; } + case IPPROTO_UDP: + if (type == SOCK_DGRAM) { af = AF_INET; break; } + default: SetLastError(WSAEPROTOTYPE); return INVALID_SOCKET; + } + + SERVER_START_REQ( create_socket ) + { + req->family = af; + req->type = type; + req->protocol = protocol; + req->access = GENERIC_READ|GENERIC_WRITE|SYNCHRONIZE; + req->flags = dwFlags; + req->inherit = TRUE; + set_error( wine_server_call( req ) ); + ret = (SOCKET)reply->handle; + } + SERVER_END_REQ; + if (ret) + { + TRACE("\tcreated %04x\n", ret ); + return ret; + } + + if (GetLastError() == WSAEACCES) /* raw socket denied */ + { + if (type == SOCK_RAW) + MESSAGE("WARNING: Trying to create a socket of type SOCK_RAW, will fail unless running as root\n"); + else + MESSAGE("WS_SOCKET: not enough privileges to create socket, try running as root\n"); + SetLastError(WSAESOCKTNOSUPPORT); + } + WARN("\t\tfailed!\n"); + return INVALID_SOCKET; +} /*********************************************************************** * __WSAFDIsSet (WINSOCK.151)

1 0

Patch: 003-sock_basic.diff
by Martin Wilck 07 Jan '02

07 Jan '02

#3 in series (incremental) Patch file: 003-sock_basic.diff Purpose: Basic implementation of socket flags. Created: 2002-01-07 Applies: CVS 2002-01-07 (with 002-fd_type.diff applied) This patch adds socket flags to the server data structure for sockets, and adds means to set these flags when creating sockets. It also provides basic server functionality to allow async I/O on sockets. ** NOTE ON BUILDING: tools/make_requests must be run ** ** after applying this patch ** server/protocol.def: create_socket request: add flags parameter. server/sock.c: struct socket: add flags field. add struct async_queue fields for async IO operations. sock_get_info(): Account for WSA_FLAG_OVERLAPPED. create_socket(): Additional parameter (unsigned int flags) Initialize async queues for overlapped sockets. create_socket_req(): Pass flags parameter to create_socket(). sock_destroy(): Destroy async queues for overlapped sockets. dlls/winsock/socket.c: WS_socket(): Initialize socket flags parameter (currently 0). Martin Wilck <Martin.Wilck(a)Fujitsu-Siemens.com> diff -ru -X diffignore CVS/wine/dlls/winsock/socket.c MW/wine/dlls/winsock/socket.c --- CVS/wine/dlls/winsock/socket.c Wed Jan 2 12:50:02 2002 +++ MW/wine/dlls/winsock/socket.c Fri Jan 4 11:13:03 2002 @@ -2381,6 +2381,7 @@ req->type = type; req->protocol = protocol; req->access = GENERIC_READ|GENERIC_WRITE|SYNCHRONIZE; + req->flags = 0U; req->inherit = TRUE; set_error( wine_server_call( req ) ); ret = (SOCKET)reply->handle; diff -ru -X diffignore CVS/wine/server/file.c MW/wine/server/file.c --- CVS/wine/server/file.c Fri Jan 4 11:07:45 2002 +++ MW/wine/server/file.c Thu Jan 3 15:21:50 2002 @@ -342,7 +342,7 @@ reply->serial = 0; /* FIXME */ } - if (file->flags & FILE_FLAG_OVERLAPPED) + if ( file->flags & FILE_FLAG_OVERLAPPED ) fd_type |= FD_TYPE_FLAG_OVERLAPPED; return fd_type; diff -ru -X diffignore CVS/wine/server/protocol.def MW/wine/server/protocol.def --- CVS/wine/server/protocol.def Fri Jan 4 11:07:45 2002 +++ MW/wine/server/protocol.def Fri Jan 4 10:20:53 2002 @@ -646,6 +646,7 @@ int family; /* family, see socket manpage */ int type; /* type, see socket manpage */ int protocol; /* protocol, see socket manpage */ + unsigned int flags; /* socket flags */ @REPLY handle_t handle; /* handle to the new socket */ @END diff -ru -X diffignore CVS/wine/server/sock.c MW/wine/server/sock.c --- CVS/wine/server/sock.c Fri Jan 4 11:07:45 2002 +++ MW/wine/server/sock.c Fri Jan 4 11:10:45 2002 @@ -36,6 +36,7 @@ #include "handle.h" #include "thread.h" #include "request.h" +#include "async.h" /* To avoid conflicts with the Unix socket headers. Plus we only need a few * macros anyway. @@ -50,8 +51,11 @@ unsigned int mask; /* event mask */ unsigned int hmask; /* held (blocked) events */ unsigned int pmask; /* pending events */ + unsigned int flags; /* socket flags */ struct event *event; /* event object */ int errors[FD_MAX_EVENTS]; /* event errors */ + struct async_queue read_q; /* Queue for asynchronous reads */ + struct async_queue write_q; /* Queue for asynchronous writes */ }; static void sock_dump( struct object *obj, int verbose ); @@ -275,6 +279,9 @@ static int sock_get_info( struct object *obj, struct get_file_info_reply *reply ) { + struct sock *sock = (struct sock*) obj; + assert ( obj->ops == &sock_ops ); + if (reply) { reply->type = FILE_TYPE_PIPE; @@ -288,6 +295,10 @@ reply->index_low = 0; reply->serial = 0; } + + if ( sock->flags & WSA_FLAG_OVERLAPPED ) + return FD_TYPE_SOCKET | FD_TYPE_FLAG_OVERLAPPED; + return FD_TYPE_SOCKET; } @@ -297,6 +308,13 @@ assert( obj->ops == &sock_ops ); /* FIXME: special socket shutdown stuff? */ + + if ( sock->flags & WSA_FLAG_OVERLAPPED ) + { + destroy_async_queue ( &sock->read_q ); + destroy_async_queue ( &sock->write_q ); + } + if (sock->event) { /* if the service thread was waiting for the event object, @@ -310,7 +328,7 @@ } /* create a new and unconnected socket */ -static struct object *create_socket( int family, int type, int protocol ) +static struct object *create_socket( int family, int type, int protocol, unsigned int flags ) { struct sock *sock; int sockfd; @@ -329,9 +347,15 @@ sock->mask = 0; sock->hmask = 0; sock->pmask = 0; + sock->flags = flags; sock->event = NULL; sock_reselect( sock ); clear_error(); + if (sock->flags & WSA_FLAG_OVERLAPPED) + { + init_async_queue (&sock->read_q); + init_async_queue (&sock->write_q); + } return &sock->obj; } @@ -377,6 +401,12 @@ acceptsock->event = NULL; if (sock->event && !(sock->mask & FD_WINE_SERVEVENT)) acceptsock->event = (struct event *)grab_object( sock->event ); + acceptsock->flags = sock->flags; + if ( acceptsock->flags & WSA_FLAG_OVERLAPPED ) + { + init_async_queue ( &acceptsock->read_q ); + init_async_queue ( &acceptsock->write_q ); + } sock_reselect( acceptsock ); clear_error(); @@ -463,7 +493,7 @@ struct object *obj; reply->handle = 0; - if ((obj = create_socket( req->family, req->type, req->protocol )) != NULL) + if ((obj = create_socket( req->family, req->type, req->protocol, req->flags )) != NULL) { reply->handle = alloc_handle( current->process, obj, req->access, req->inherit ); release_object( obj );

1 0

Patch: 002-fd_type.diff
by Martin Wilck 07 Jan '02

07 Jan '02

Patch 2 in the series, a repost of my patch to wine-devel of 2002-01-03, Now incremental to the previous patch (i.e., also relies on Mike's patches being applied). Patch file: 002-fd_type.diff Purpose: More general approach to file descriptor types and flags. Created: 2002-01-07 Applies: CVS 2002-01-07 (with 001-file_overlapped.diff applied) If overlapped IO is to be allowed / implemented for several different types of file descriptors (regular files, serial ports, pipes, sockets, ...), a more general scheme must be found to obtain file descriptor types and flags or capabilities. The current scheme treats FD_TYPE_OVERLAPPED as a special type, making it impossible to distinguish between e.g. overlapped files and overlapped sockets. This patch fixes this behaviour by modifying the meaning of return value of the get_file_info() function. ** NOTE ON BUILDING: tools/make_requests must be run ** ** after applying this patch ** server/protocol.def: The lowest byte of the type value is reserved for the type itself (macro FD_TYPE (type)the rest is available for miscellaneous flags. New FD types: FD_TYPE_DEVICE, FD_TYPE_NAMED_PIPE, FD_TYPE_PIPE, FD_TYPE_SERIAL, FD_TYPE_SOCKET. New FD flags (were formerly FD types): FD_TYPE_FLAG_OVERLAPPED: The fd uses overlapped I/O. FD_TYPE_FLAG_TIMEOUT: The fd applies a timeout on its I/O operations New test macros: FD_TYPE_IS_OVERLAPPED (type), FD_TYPE_HAS_TIMEOUT (type) server/device.c: device_get_info(): return FD_TYPE_DEVICE. server/file.c: file_get_info(): set FD_TYPE_FLAG_OVERLAPPED on overlapped files. server/named_pipe.c: named_pipe_get_info(): return FD_TYPE_NAMED_PIPE. server/pipe.c: named_pipe_get_info(): return FD_TYPE_PIPE. server/serial.c: serial_get_info(): return FD_TYPE_SERIAL, account for OVERLAPPED and TIMEOUT flags. server/serial.c: sock_get_info(): return FD_TYPE_SOCKET. files/file.c: ReadFile(): Replace FD_TYPE_XYZ dependent switch statement with an equivalent if/elseif construct which will allow a morte fine-grained destinction of fd types and flags in the future. WriteFile(): dito. Martin Wilck <Martin.Wilck(a)Fujitsu-Siemens.com> diff -ruX diffignore CVS/wine/files/file.c MW/wine/files/file.c --- CVS/wine/files/file.c Mon Jan 7 16:04:31 2002 +++ MW/wine/files/file.c Mon Jan 7 16:00:33 2002 @@ -1467,9 +1467,8 @@ unix_handle = FILE_GetUnixHandleType( hFile, GENERIC_READ, &type ); - switch (type) + if ( FD_TYPE_IS_OVERLAPPED (type) ) { - case FD_TYPE_OVERLAPPED: if (unix_handle == -1) return FALSE; if ( (overlapped==NULL) || NtResetEvent( overlapped->hEvent, NULL ) ) { @@ -1512,15 +1511,21 @@ /* fail on return, with ERROR_IO_PENDING */ SetLastError(ERROR_IO_PENDING); return FALSE; + } - case FD_TYPE_CONSOLE: + else if ( FD_TYPE (type) == FD_TYPE_CONSOLE ) + { return ReadConsoleA(hFile, buffer, bytesToRead, bytesRead, NULL); + } - case FD_TYPE_TIMEOUT: + else if ( FD_TYPE_HAS_TIMEOUT (type) ) + { close(unix_handle); return FILE_TimeoutRead(hFile, buffer, bytesToRead, bytesRead); + } - default: + else + { /* normal unix files */ if (unix_handle == -1) return FALSE; @@ -1530,7 +1535,6 @@ SetLastError(ERROR_INVALID_PARAMETER); return FALSE; } - break; } /* code for synchronous reads */ @@ -1689,9 +1693,8 @@ unix_handle = FILE_GetUnixHandleType( hFile, GENERIC_WRITE, &type ); - switch (type) + if ( FD_TYPE_IS_OVERLAPPED (type) ) { - case FD_TYPE_OVERLAPPED: if (unix_handle == -1) return FALSE; if ( (overlapped==NULL) || NtResetEvent( overlapped->hEvent, NULL ) ) { @@ -1741,11 +1744,17 @@ SetLastError(ERROR_IO_PENDING); return FALSE; - case FD_TYPE_CONSOLE: + } + + else if ( FD_TYPE (type) == FD_TYPE_CONSOLE ) + { TRACE("%d %s %ld %p %p\n", hFile, debugstr_an(buffer, bytesToWrite), bytesToWrite, bytesWritten, overlapped ); return WriteConsoleA(hFile, buffer, bytesToWrite, bytesWritten, NULL); - default: + } + + else + { if (unix_handle == -1) return FALSE; } diff -ruX diffignore CVS/wine/server/device.c MW/wine/server/device.c --- CVS/wine/server/device.c Mon Jan 7 15:45:26 2002 +++ MW/wine/server/device.c Mon Jan 7 16:00:33 2002 @@ -83,7 +83,7 @@ reply->index_low = 0; reply->serial = 0; } - return FD_TYPE_DEFAULT; + return FD_TYPE_DEVICE; } /* create a device */ diff -ruX diffignore CVS/wine/server/file.c MW/wine/server/file.c --- CVS/wine/server/file.c Mon Jan 7 15:58:22 2002 +++ MW/wine/server/file.c Mon Jan 7 16:00:33 2002 @@ -308,6 +308,7 @@ { struct stat st; struct file *file = (struct file *)obj; + int fd_type = FD_TYPE_DEFAULT; assert( obj->ops == &file_ops ); if (reply) @@ -342,9 +343,10 @@ } if (file->flags & FILE_FLAG_OVERLAPPED) - return FD_TYPE_OVERLAPPED; + fd_type |= FD_TYPE_FLAG_OVERLAPPED; + + return fd_type; - return FD_TYPE_DEFAULT; } static struct async_queue *file_queue_async(struct object *obj, struct async *async, int type, int count) diff -ruX diffignore CVS/wine/server/named_pipe.c MW/wine/server/named_pipe.c --- CVS/wine/server/named_pipe.c Mon Jan 7 15:45:26 2002 +++ MW/wine/server/named_pipe.c Mon Jan 7 16:00:33 2002 @@ -197,7 +197,7 @@ reply->index_low = 0; reply->serial = 0; } - return FD_TYPE_DEFAULT; + return FD_TYPE_NAMED_PIPE; } static struct named_pipe *create_named_pipe( const WCHAR *name, size_t len ) diff -ruX diffignore CVS/wine/server/pipe.c MW/wine/server/pipe.c --- CVS/wine/server/pipe.c Mon Jan 7 15:45:26 2002 +++ MW/wine/server/pipe.c Mon Jan 7 16:00:33 2002 @@ -140,7 +140,7 @@ reply->index_low = 0; reply->serial = 0; } - return FD_TYPE_DEFAULT; + return FD_TYPE_PIPE; } static void pipe_destroy( struct object *obj ) diff -ruX diffignore CVS/wine/server/protocol.def MW/wine/server/protocol.def --- CVS/wine/server/protocol.def Mon Jan 7 15:45:26 2002 +++ MW/wine/server/protocol.def Mon Jan 7 16:00:33 2002 @@ -544,12 +544,22 @@ int fd; /* file descriptor */ int type; /* the type of file */ @END +#define FD_TYPE(x) ((x) & 0xff) + #define FD_TYPE_INVALID 0 #define FD_TYPE_DEFAULT 1 #define FD_TYPE_CONSOLE 2 -#define FD_TYPE_OVERLAPPED 3 -#define FD_TYPE_TIMEOUT 4 +#define FD_TYPE_DEVICE 3 +#define FD_TYPE_NAMED_PIPE 4 +#define FD_TYPE_PIPE 5 +#define FD_TYPE_SERIAL 6 +#define FD_TYPE_SOCKET 7 + +#define FD_TYPE_FLAG_OVERLAPPED 0x100 +#define FD_TYPE_FLAG_TIMEOUT 0x200 +#define FD_TYPE_IS_OVERLAPPED(x) ((x) & FD_TYPE_FLAG_OVERLAPPED) +#define FD_TYPE_HAS_TIMEOUT(x) ((x) & FD_TYPE_FLAG_TIMEOUT) /* Set a file current position */ @REQ(set_file_pointer) diff -ruX diffignore CVS/wine/server/serial.c MW/wine/server/serial.c --- CVS/wine/server/serial.c Mon Jan 7 15:45:26 2002 +++ MW/wine/server/serial.c Mon Jan 7 16:00:33 2002 @@ -192,6 +192,7 @@ static int serial_get_info( struct object *obj, struct get_file_info_reply *reply ) { struct serial *serial = (struct serial *) obj; + int fd_type = FD_TYPE_SERIAL; assert( obj->ops == &serial_ops ); if (reply) @@ -209,13 +210,13 @@ } if(serial->attrib & FILE_FLAG_OVERLAPPED) - return FD_TYPE_OVERLAPPED; + fd_type |= FD_TYPE_FLAG_OVERLAPPED; - if( (serial->readinterval == MAXDWORD) && - (serial->readmult == 0) && (serial->readconst == 0) ) - return FD_TYPE_DEFAULT; + else if( !((serial->readinterval == MAXDWORD) && + (serial->readmult == 0) && (serial->readconst == 0)) ) + fd_type |= FD_TYPE_FLAG_TIMEOUT; - return FD_TYPE_TIMEOUT; + return fd_type; } static void serial_poll_event(struct object *obj, int event) diff -ruX diffignore CVS/wine/server/sock.c MW/wine/server/sock.c --- CVS/wine/server/sock.c Mon Jan 7 15:45:26 2002 +++ MW/wine/server/sock.c Mon Jan 7 16:00:33 2002 @@ -288,7 +288,7 @@ reply->index_low = 0; reply->serial = 0; } - return FD_TYPE_DEFAULT; + return FD_TYPE_SOCKET; } static void sock_destroy( struct object *obj )

1 0

Patch: 001-file_overlapped.diff
by Martin Wilck 07 Jan '02

07 Jan '02

This patch is a repost of my patch of 2002-01-03, now against the current CVS version including Mike McCormack's latest patches (01-cancelio.diff, 01-write_file_update.diff, 01-comm16_update.diff as of 2002-01-06). It is the first in a series of patches that constitute intermediate steps towards async socket IO. I consider them pretty clean and have done basic testing. My testing possibilities are limited, though - I'd be grateful to anybody who could run tests especially with 16-bit network-oriented windows applications. Bugs in my code should show up immediately as regressions. Patch file: 001-file_overlapped.diff Purpose: This patch fixes overlapped I/O on regular files. Created: 2002-01-07 Applies: CVS 2002-01-07 with Mike McCorkack's patches of 2002-01-06 applied (01-cancelio.diff, 01-write_file_update.diff, 01-comm16_update.diff) Add server side support for async IO on files. FILE_Async[Read|Write]Service and WriteFile() use pread() / pwrite() rather than read() / write () for fd's on which lseek()ing is possible to honour the fact that lpOverlapped->Offset and lpOverlapped->OffsetHigh contain the offset from which the file is to be read. files/file.c: OVERLAPPED_OFFSET: Macro to calculate file offset from OVERLAPPED struct. FILE_AsyncReadService(): Use pread() on fd's that allow lseek(). FILE_AsyncWriteService(): Use pwrite() on fd's that allow lseek(). WriteFile(): dito. server/file.c: struct file: Add two struct async_queue fields (for async read/write). file_ops: file specific poll_event() method accounts for overlapped IO. define queue_async() method. create_file_for_fd(): call init_async_queue() on overlapped files. file_poll_event(): call default_poll_event() only after checking for queued async requests. file_get_info(): return FD_TYPE_OVERLAPPED if FILE_FLAG_OVERLAPPED is set. file_queue_async(): implement queue_async() method for files. file_destroy(): destroy async queues on overlapped files. Martin Wilck <Martin.Wilck(a)Fujitsu-Siemens.com> diff -ruX diffignore CVS/wine/files/file.c MW/wine/files/file.c --- CVS/wine/files/file.c Mon Jan 7 15:54:59 2002 +++ MW/wine/files/file.c Mon Jan 7 15:55:36 2002 @@ -9,6 +9,9 @@ * Right now, they simply call the CopyFile method. */ +/* This is needed to avoid compile-time warnings about pread / pwrite */ +#define _GNU_SOURCE +#define _XOPEN_SOURCE 500 #include "config.h" #include "wine/port.h" @@ -55,6 +58,9 @@ /* Size of per-process table of DOS handles */ #define DOS_TABLE_SIZE 256 +/* Macro to derive file offset from OVERLAPPED struct */ +#define OVERLAPPED_OFFSET(overlapped) ((off_t) (overlapped)->Offset + ((off_t) (overlapped)->OffsetHigh << 32)) + static HANDLE dos_handles[DOS_TABLE_SIZE]; @@ -1304,12 +1310,21 @@ { LPOVERLAPPED lpOverlapped = ovp->lpOverlapped; int result, r; + int already = lpOverlapped->InternalHigh; TRACE("%p %p\n", lpOverlapped, ovp->buffer ); /* check to see if the data is ready (non-blocking) */ - result = read(ovp->fd, &ovp->buffer[lpOverlapped->InternalHigh], - ovp->count - lpOverlapped->InternalHigh); + + /* Check if this file object can lseek() */ + /* FIXME: test this only once for each file object ?? */ + if (lseek (ovp->fd, 0, SEEK_CUR) != (off_t) -1) + result = pread (ovp->fd, + &ovp->buffer[already], + ovp->count - already, + OVERLAPPED_OFFSET (lpOverlapped) + already); + else + result = read (ovp->fd, &ovp->buffer[already], ovp->count - already); if ( (result<0) && ((errno == EAGAIN) || (errno == EINTR))) { @@ -1543,12 +1558,21 @@ { LPOVERLAPPED lpOverlapped = ovp->lpOverlapped; int result, r; + int already = lpOverlapped->InternalHigh; TRACE("(%p %p)\n",lpOverlapped,ovp->buffer); /* write some data (non-blocking) */ - result = write(ovp->fd, &ovp->buffer[lpOverlapped->InternalHigh], - ovp->count-lpOverlapped->InternalHigh); + + /* Check if this file object can lseek() */ + /* FIXME: test this only once for each file object ?? */ + if (lseek (ovp->fd, 0, SEEK_CUR) != (off_t) -1) + result = pwrite (ovp->fd, + &ovp->buffer[already], + ovp->count - already, + OVERLAPPED_OFFSET (lpOverlapped) + already); + else + result = write (ovp->fd, &ovp->buffer[already], ovp->count - already); if ( (result<0) && ((errno == EAGAIN) || (errno == EINTR))) { @@ -1678,7 +1702,13 @@ } /* see if we can write some data already (this shouldn't block) */ - result = write( unix_handle, buffer, bytesToWrite ); + + if ( lseek(unix_handle, 0, SEEK_CUR) != (off_t) -1 ) + result = pwrite( unix_handle, buffer, bytesToWrite, + OVERLAPPED_OFFSET (overlapped) ); + else + result = write( unix_handle, buffer, bytesToWrite ); + close(unix_handle); if(result<0) @@ -1693,7 +1723,7 @@ result = 0; } - /* if we read enough to keep the app happy, then return now */ + /* if we wrote enough to keep the app happy, then return now */ if(result>=bytesToWrite) { *bytesWritten = result; diff -ruX diffignore CVS/wine/server/file.c MW/wine/server/file.c --- CVS/wine/server/file.c Mon Jan 7 15:54:59 2002 +++ MW/wine/server/file.c Mon Jan 7 15:55:28 2002 @@ -29,6 +29,7 @@ #include "handle.h" #include "thread.h" #include "request.h" +#include "async.h" struct file { @@ -39,6 +40,9 @@ unsigned int flags; /* flags (FILE_FLAG_*) */ unsigned int sharing; /* file sharing mode */ int drive_type; /* type of drive the file is on */ + + struct async_queue read_q; + struct async_queue write_q; }; #define NAME_HASH_SIZE 37 @@ -47,10 +51,12 @@ static void file_dump( struct object *obj, int verbose ); static int file_get_poll_events( struct object *obj ); +static void file_poll_event( struct object *obj, int event ); static int file_get_fd( struct object *obj ); static int file_flush( struct object *obj ); static int file_get_info( struct object *obj, struct get_file_info_reply *reply ); static void file_destroy( struct object *obj ); +static struct async_queue * file_queue_async(struct object *obj, struct async* async, int type, int count); static const struct object_ops file_ops = { @@ -61,11 +67,11 @@ default_poll_signaled, /* signaled */ no_satisfied, /* satisfied */ file_get_poll_events, /* get_poll_events */ - default_poll_event, /* poll_event */ + file_poll_event, /* poll_event */ file_get_fd, /* get_fd */ file_flush, /* flush */ file_get_info, /* get_file_info */ - NULL, /* queue_async */ + file_queue_async, /* queue_async */ file_destroy /* destroy */ }; @@ -116,6 +122,11 @@ file->flags = attrs; file->sharing = sharing; file->drive_type = drive_type; + if (file->flags & FILE_FLAG_OVERLAPPED) + { + init_async_queue (&file->read_q); + init_async_queue (&file->write_q); + } } return file; } @@ -253,6 +264,27 @@ return events; } +static void file_poll_event( struct object *obj, int event ) +{ + struct file *file = (struct file *)obj; + assert( obj->ops == &file_ops ); + if ( file->flags & FILE_FLAG_OVERLAPPED ) + { + if( IS_READY(file->read_q) && (POLLIN & event) ) + { + async_notify(file->read_q.head, STATUS_ALERTED); + return; + } + if( IS_READY(file->write_q) && (POLLIN & event) ) + { + async_notify(file->write_q.head, STATUS_ALERTED); + return; + } + } + default_poll_event( obj, event ); +} + + static int file_get_fd( struct object *obj ) { struct file *file = (struct file *)obj; @@ -308,9 +340,45 @@ reply->index_low = st.st_ino; reply->serial = 0; /* FIXME */ } + + if (file->flags & FILE_FLAG_OVERLAPPED) + return FD_TYPE_OVERLAPPED; + return FD_TYPE_DEFAULT; } +static struct async_queue *file_queue_async(struct object *obj, struct async *async, int type, int count) +{ + struct file *file = (struct file *)obj; + struct async_queue *q; + + assert( obj->ops == &file_ops ); + + if ( !(file->flags & FILE_FLAG_OVERLAPPED) ) + { + set_error ( STATUS_INVALID_HANDLE ); + return NULL; + } + + switch(type) + { + case ASYNC_TYPE_READ: + q = &file->read_q; + break; + case ASYNC_TYPE_WRITE: + q = &file->write_q; + break; + default: + set_error( STATUS_INVALID_PARAMETER ); + return NULL; + } + + if(async && !async->q) + async_insert(q, async); + + return q; +} + static void file_destroy( struct object *obj ) { struct file *file = (struct file *)obj; @@ -325,6 +393,11 @@ *pptr = (*pptr)->next; if (file->flags & FILE_FLAG_DELETE_ON_CLOSE) unlink( file->name ); free( file->name ); + } + if (file->flags & FILE_FLAG_OVERLAPPED) + { + destroy_async_queue (&file->read_q); + destroy_async_queue (&file->write_q); } }

1 0

mvcrt patch
by Waldek Hebisch 07 Jan '02

07 Jan '02

Enclosed is a patch (diff to 2001.12.26 snapshot) which impements stdio buffering in msvcrt. Since I cannot make Visual C binaries could anybady with Visual C++ test it? I have testet it with mingw32 and I incude my test program (iotst.c). The program prints a line indicating which test it is making and then complains about failures. Comment 1: current builtin msvcrt fails test1 (and few other as a consequence). Some native versions of msvcrt fail test 8. Comment 2: I wrote the test the way naive programmer would wrote it (IMHO). Now I think that for a testing framework we need an assert-like macro to make checking and printing code shorter and to have uniform printouts. I ommited some tests that require multiple processes (expandig file while it is read, checking that stdio buffers got flushed at exit) - I guess that doing such test in portable way is quite tricky. -- Waldek Hebisch hebisch(a)math.uni.wroc.pl or hebisch(a)hera.math.uni.wroc.pl

1 0

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

wine-devel