diff options
Diffstat (limited to 'gdb/osf-share/AT386')
| -rw-r--r-- | gdb/osf-share/AT386/.Sanitize | 44 | ||||
| -rw-r--r-- | gdb/osf-share/AT386/cma_thread_io.h | 457 |
2 files changed, 501 insertions, 0 deletions
diff --git a/gdb/osf-share/AT386/.Sanitize b/gdb/osf-share/AT386/.Sanitize new file mode 100644 index 0000000..2ef8e54 --- /dev/null +++ b/gdb/osf-share/AT386/.Sanitize @@ -0,0 +1,44 @@ +# .Sanitize for devo/gdb/osf-share/AT386. + +# Each directory to survive its way into a release will need a file +# like this one called "./.Sanitize". All keyword lines must exist, +# and must exist in the order specified by this file. Each directory +# in the tree will be processed, top down, in the following order. + +# Hash started lines like this one are comments and will be deleted +# before anything else is done. Blank lines will also be squashed +# out. + +# The lines between the "Do-first:" line and the "Things-to-keep:" +# line are executed as a /bin/sh shell script before anything else is +# done in this directory. + +Do-first: + +# All files listed between the "Things-to-keep:" line and the +# "Files-to-sed:" line will be kept. All other files will be removed. +# Directories listed in this section will have their own Sanitize +# called. Directories not listed will be removed in their entirety +# with rm -rf. + +Things-to-keep: + +cma_thread_io.h + +# Things which are explicitly *not* kept, for now. + +Things-to-lose: + +Do-last: + +# Don't try to clean directories here, as the 'mv' command will fail. +# Also, grep fails on NFS mounted directories. + +for i in * ; do + if test ! -d $i && (grep sanitize $i > /dev/null) ; then + echo '***' Some mentions of Sanitize are still left in $i! 1>&2 + fi +done + +# +# End of file. diff --git a/gdb/osf-share/AT386/cma_thread_io.h b/gdb/osf-share/AT386/cma_thread_io.h new file mode 100644 index 0000000..a90dba1 --- /dev/null +++ b/gdb/osf-share/AT386/cma_thread_io.h @@ -0,0 +1,457 @@ +/* + * (c) Copyright 1990-1996 OPEN SOFTWARE FOUNDATION, INC. + * (c) Copyright 1990-1996 HEWLETT-PACKARD COMPANY + * (c) Copyright 1990-1996 DIGITAL EQUIPMENT CORPORATION + * (c) Copyright 1991, 1992 Siemens-Nixdorf Information Systems + * To anyone who acknowledges that this file is provided "AS IS" without + * any express or implied warranty: permission to use, copy, modify, and + * distribute this file for any purpose is hereby granted without fee, + * provided that the above copyright notices and this notice appears in + * all source code copies, and that none of the names listed above be used + * in advertising or publicity pertaining to distribution of the software + * without specific, written prior permission. None of these organizations + * makes any representations about the suitability of this software for + * any purpose. + */ +/* + * Header file for thread synchrounous I/O + */ + +#ifndef CMA_THREAD_IO +#define CMA_THREAD_IO + +/* + * INCLUDE FILES + */ + +#include <cma_config.h> +#include <sys/file.h> +#include <cma.h> +#include <sys/types.h> +#include <sys/time.h> +#include <cma_init.h> +#include <cma_errors.h> + +/* + * CONSTANTS + */ + +/* + * Define symbols which indicate whether to compile code for obsolete + * "non-blocking mode" flags: FNDELAY and FNBLOCK. If the obsolete + * symbols are defined, and if their replacement symbols are defined + * and are different or if they are undefined, then define a symbol + * that says to compile the code in; otherwise no code will be compiled + * for these obsolete symbols. + */ +#ifdef FNDELAY +# ifdef O_NDELAY +# if O_NDELAY != FNDELAY +# define _CMA_FNDELAY_ +# endif +# else +# define _CMA_FNDELAY_ +# endif +#endif + +#ifdef FNBLOCK +# ifdef O_NONBLOCK +# if O_NONBLOCK != FNBLOCK +# define _CMA_FNBLOCK_ +# endif +# else +# define _CMA_FNBLOCK_ +# endif +#endif + + +extern cma_t_boolean cma_is_open(int); +/* + * Maximum number of files (ie, max_fd+1) + */ +#define cma__c_mx_file FD_SETSIZE + +/* + * Number of bits per file descriptor bit mask (ie number of bytes * bits/byte) + */ +#define cma__c_nbpm NFDBITS + +/* + * TYPE DEFINITIONS + */ + +typedef enum CMA__T_IO_TYPE { + cma__c_io_read = 0, + cma__c_io_write = 1, + cma__c_io_except = 2 + } cma__t_io_type; +#define cma__c_max_io_type 2 + +/* + * From our local <sys/types.h>: + * + * typedef long fd_mask; + * + * typedef struct fd_set { + * fd_mask fds_bits[howmany(FD_SETSIZE, NFDBITS)]; + * } fd_set; + * + */ +typedef fd_mask cma__t_mask; +typedef fd_set cma__t_file_mask; + + +/* + * GLOBAL DATA + */ + +/* + * Maximum number of files (ie, max_fd+1) as determined by getdtablesize(). + */ +extern int cma__g_mx_file; + +/* + * Number of submasks (ie "int" sized chunks) per file descriptor mask as + * determined by getdtablesize(). + */ +extern int cma__g_nspm; + +/* + * MACROS + */ + +/* + * Define a constant for the errno value which indicates that the requested + * operation was not performed because it would block the process. + */ +# define cma__is_blocking(s) \ + ((s == EAGAIN) || (s == EWOULDBLOCK) || (s == EINPROGRESS) || \ + (s == EALREADY) || (s == EDEADLK)) + +/* +* It is necessary to issue an I/O function, before calling cma__io_wait() +* in the following cases: +* +* * This file descriptor has been set non-blocking by CMA +* * This file descriptor has been set non-blocking by the user. +*/ + +#define cma__issue_io_call(fd) \ + ( (cma__g_file[fd]->non_blocking) || \ + (cma__g_file[fd]->user_fl.user_non_blocking) ) + + +#define cma__set_user_nonblocking(flags) \ + +/* + * Determine if the file is open + */ +/* + * If the file gets closed while waiting for the mutex cma__g_file[rfd] + * gets set to null. This results in a crash if NDEBUG is set to 0 + * since cma__int_lock tries to dereference it to set the mutex ownership + * after it gets the mutex. The following will still set the ownership + * in cma__int_lock so we'll set it back to noone if cma__g_file is null + * when we come back just in case it matters. It shouldn't since its no + * longer in use but..... + * Callers of this should recheck cma__g_file after the reservation to + * make sure continueing makes sense. + */ +#define cma__fd_reserve(rfd) \ + { \ + cma__t_int_mutex *__mutex__; \ + __mutex__ = cma__g_file[rfd]->mutex; \ + cma__int_lock (__mutex__); \ + if(cma__g_file[rfd] == (cma__t_file_obj *)cma_c_null_ptr) \ + cma__int_unlock(__mutex__); \ + } + + +/* + * Unreserve a file descriptor + */ +#define cma__fd_unreserve(ufd) cma__int_unlock (cma__g_file[ufd]->mutex) + +/* + * AND together two select file descriptor masks + */ +#define cma__fdm_and(target,a,b) \ + { \ + int __i__ = cma__g_nspm; \ + while (__i__--) \ + (target)->fds_bits[__i__] = \ + (a)->fds_bits[__i__] & (b)->fds_bits[__i__]; \ + } + +/* + * Clear a bit in a select file descriptor mask + * + * FD_CLR(n, p) := ((p)->fds_bits[(n)/NFDBITS] &= ~(1 << ((n) % NFDBITS))) + */ +#define cma__fdm_clr_bit(n,p) FD_CLR (n, p) + +/* + * Copy the contents of one file descriptor mask into another. If the + * destination operand is null, do nothing; if the source operand is null, + * simply zero the destination. + */ +#define cma__fdm_copy(src,dst,nfds) { \ + if (dst) \ + if (src) { \ + cma__t_mask *__s__ = (cma__t_mask *)(src); \ + cma__t_mask *__d__ = (cma__t_mask *)(dst); \ + int __i__; \ + for (__i__ = 0; __i__ < (nfds); __i__ += cma__c_nbpm) \ + *__d__++ = *__s__++; \ + } \ + else \ + cma__fdm_zero (dst); \ + } + +/* + * To increment count for each bit set in fd - mask + */ +#define cma__fdm_count_bits(map,count) \ + { \ + int __i__ = cma__g_nspm; \ + while (__i__--) { \ + cma__t_mask __tm__; \ + __tm__ = (map)->fds_bits[__i__]; \ + while(__tm__) { \ + (count)++; \ + __tm__ &= ~(__tm__ & (-__tm__)); /* Assumes 2's comp */ \ + } \ + } \ + } + +/* + * Test if a bit is set in a select file descriptor mask + * + * FD_ISSET(n,p) := ((p)->fds_bits[(n)/NFDBITS] & (1 << ((n) % NFDBITS))) + */ +#define cma__fdm_is_set(n,p) FD_ISSET (n, p) + +/* + * OR together two select file descriptor masks + */ +#define cma__fdm_or(target,a,b) \ + { \ + int __i__ = cma__g_nspm; \ + while (__i__--) \ + (target)->fds_bits[__i__] = \ + (a)->fds_bits[__i__] | (b)->fds_bits[__i__]; \ + } + +/* + * Set a bit in a select file descriptor mask + * + * FD_SET(n,p) := ((p)->fds_bits[(n)/NFDBITS] |= (1 << ((n) % NFDBITS))) + */ +#define cma__fdm_set_bit(n,p) FD_SET (n, p) + +/* + * Clear a select file descriptor mask. + */ +#define cma__fdm_zero(n) \ + cma__memset ((char *) n, 0, cma__g_nspm * sizeof(cma__t_mask)) + + + +/* + * CMA "thread-synchronous" I/O read/write operations + */ + + /* + * Since all CMA "thread-synchronous" I/O (read or write) operations on + * U*ix follow the exact same structure, the wrapper routines have been + * condensed into a macro. + * + * The steps performed are as follows: + * 1. Check that the file descriptor is a legitimate value. + * 2. Check that the entry in the CMA file "database" which corresponds to + * the file descriptor indicates that the "file" was "opened" by CMA. + * 3. Reserve the file, to serialized access to files. This not only + * simplifies things, but also defends against non-reentrancy. + * 4. If the "file" is "set" for non-blocking I/O, check if we + * have actually set the file non-blocking yet, and if not do so. + * Then, issue the I/O operantion. + * Success or failure is returned immediately, after unreserving the + * file. If the error indicates that the operation would have caused + * the process to block, continue to the next step. + * 5. The I/O prolog adds this "file" to the global bit mask, which + * represents all "files" which have threads waiting to perform I/O on + * them, and causes the thread to block on the condition variable for + * this "file". Periodically, a select is done on this global bit + * mask, and the condition variables corresponding to "files" which + * are ready for I/O are signaled, releasing those waiting threads to + * perform their I/O. + * 6. When the thread returns from the I/O prolog, it can (hopefully) + * perform its operation without blocking the process. + * 7. The I/O epilog clears the bit in the global mask and/or signals the + * the next thread waiting for this "file", as appropriate. + * 8. If the I/O failed, continue to loop. + * 9. Finally, the "file" is unreserved, as we're done with it, and the + * result of the operation is returned. + * + * + * Note: currently, we believe that timeslicing which is based on the + * virtual-time timer does not cause system calls to return EINTR. + * Threfore, any EINTR returns are relayed directly to the caller. + * On platforms which do not support a virtual-time timer, the code + * should probably catch EINTR returns and restart the system call. + */ + +/* + * This macro is used for both read-type and write-type functions. + * + * Note: the second call to "func" may require being bracketed in a + * cma__interrupt_disable/cma__interrupt_enable pair, but we'll + * wait and see if this is necessary. + */ +#define cma__ts_func(func,fd,arglist,type,post_process) { \ + cma_t_integer __res__; \ + cma_t_boolean __done__ = cma_c_false; \ + if ((fd < 0) || (fd >= cma__g_mx_file)) return (cma__set_errno (EBADF), -1); \ + if (!cma__is_open(fd)) return (cma__set_errno (EBADF), -1); \ + cma__fd_reserve (fd); \ + if (!cma__is_open(fd)) return (cma__set_errno (EBADF), -1); \ + if (cma__issue_io_call(fd)) {\ + if ((!cma__g_file[fd]->set_non_blocking) && \ + (cma__g_file[fd]->non_blocking == cma_c_true)) \ + cma__set_nonblocking(fd); \ + cma__interrupt_disable (0); \ + TRY { \ + __res__ = func arglist; \ + } \ + CATCH_ALL { \ + cma__interrupt_enable (0); \ + cma__fd_unreserve (fd); \ + RERAISE; \ + } \ + ENDTRY \ + cma__interrupt_enable (0); \ + if ((__res__ != -1) \ + || (!cma__is_blocking (errno)) \ + || (cma__g_file[fd]->user_fl.user_non_blocking)) \ + __done__ = cma_c_true; \ + } \ + if (__done__) { \ + cma__fd_unreserve (fd); \ + } \ + else { \ + TRY { \ + cma__io_prolog (type, fd); \ + while (!__done__) { \ + cma__io_wait (type, fd); \ + __res__ = func arglist; \ + if ((__res__ != -1) \ + || (!cma__is_blocking (errno)) \ + || (cma__g_file[fd]->user_fl.user_non_blocking)) \ + __done__ = cma_c_true; \ + } \ + } \ + FINALLY { \ + cma__io_epilog (type, fd); \ + cma__fd_unreserve (fd); \ + } \ + ENDTRY \ + } \ + if (__res__ != -1) post_process; \ + return __res__; \ + } + + /* + * Since most CMA "thread-synchronous" I/O ("open"-type) operations on + * U*ix follow the exact same structure, the wrapper routines have been + * condensed into a macro. + * + * The steps performed are as follows: + * 1. Issue the open function. + * 2. If the value returned indicates an error, return it to the caller. + * 3. If the file descriptor returned is larger than what we think is the + * maximum value (ie if it is too big for our database) then bugcheck. + * 4. "Open" the "file" in the CMA file database. + * 5. Return the file descriptor value to the caller. + * + * FIX-ME: for the time being, if the I/O operation returns EINTR, we + * simply return it to the caller; eventually, we should catch this + * and "do the right thing" (if we can figure out what that is). + */ + +/* + * This macro is used for all "open"-type functions which return a single file + * desciptor by immediate value. + */ +#define cma__ts_open(func,arglist,post_process) { \ + int __fd__; \ + TRY { \ + cma__int_init (); \ + cma__int_lock (cma__g_io_data_mutex); \ + __fd__ = func arglist; \ + cma__int_unlock (cma__g_io_data_mutex); \ + if (__fd__ >= 0 && __fd__ < cma__g_mx_file) \ + post_process; \ + } \ + CATCH_ALL \ + { \ + cma__set_errno (EBADF); \ + __fd__ = -1; \ + } \ + ENDTRY \ + if (__fd__ >= cma__g_mx_file) \ + cma__bugcheck ("cma__ts_open: fd is too large"); \ + return __fd__; \ + } +/* + * This macro is used for all "open"-type functions which return a pair of file + * desciptors by reference parameter. + */ +#define cma__ts_open2(func,fdpair,arglist,post_process) { \ + int __res__; \ + TRY { \ + cma__int_init (); \ + cma__int_lock (cma__g_io_data_mutex); \ + __res__ = func arglist; \ + cma__int_unlock (cma__g_io_data_mutex); \ + if (__res__ >= 0 && fdpair[0] < cma__g_mx_file \ + && fdpair[1] < cma__g_mx_file) \ + post_process; \ + } \ + CATCH_ALL \ + { \ + cma__set_errno (EBADF); \ + __res__ = -1; \ + } \ + ENDTRY \ + if ((fdpair[0] >= cma__g_mx_file) || (fdpair[1] >= cma__g_mx_file)) \ + cma__bugcheck ("cma__ts_open2: one of fd's is too large"); \ + return __res__; \ + } + +/* + * INTERNAL INTERFACES + */ +extern void cma__close_general (int); + +extern void cma__init_thread_io (void); + +extern cma_t_boolean cma__io_available (cma__t_io_type,int,struct timeval *); + +extern void cma__io_epilog (cma__t_io_type,int); + +extern void cma__io_prolog (cma__t_io_type,int); + +extern void cma__io_wait (cma__t_io_type,int); + +extern void cma__open_general (int); + +extern void cma__reinit_thread_io (int); + +extern void cma__set_nonblocking (int); + +extern void cma__set_user_nonblock_flags (int,int); + +extern cma_t_boolean cma__is_open (int); + + +#endif |