rayan@cs.toronto.edu (Rayan Zachariassen) (07/28/89)
q1: (curiosity) How can I reliably determine the end of the envp strings in the process address space? I.e. what is between the last environment string value and stackbas(0) (from core.h) ? q2: (problem) What magic do I need to read the top of user stack through /debug/<pid>? I tried the following sequence: fd = open("/debug/<pid>", 0); fcntl(fd, DFCSTOP, 0); nfd = fcntl(fd, DFCOPENT, x); lseek(nfd, stackbas(some low number), 0) read(nfd, buf, sizeof buf) fcntl(fd, DFCRUN, &(something which is CLEARNOSIG)) with x being variously 0, i, and &i, where i=0 or i=stackbas(some low number) without luck. I also tried seeking and reading from fd, but got something that looks like initialized data and/or symbol table. stackbas(some low number) ~= 0x7fffc700 q3: (curiosity) What is the reason to require the process to be stopped before doing DFCOPENT or read() from it? This limits the usefulness of the operations provided. My purpose is to rummage through the inherited environment of an arbitrary, running, possibly critical, process (like login shells...). Thanks for any info rayan
jeffd@norge.sgi.com (Jeff Doughty) (07/29/89)
I can answer two of the three questions. Perhaps one of my colleagues can answer #1. > q2: (problem) What magic do I need to read the top of user stack through > /debug/<pid>? I tried the following sequence: > > fd = open("/debug/<pid>", 0); > fcntl(fd, DFCSTOP, 0); > nfd = fcntl(fd, DFCOPENT, x); > lseek(nfd, stackbas(some low number), 0) > read(nfd, buf, sizeof buf) > fcntl(fd, DFCRUN, &(something which is CLEARNOSIG)) This is almost correct. The problem is that you are reading the wrong file descriptor. DFCOPENT returns the file descriptor of the file that contains the binary. For example, if you are running "/usr/tmp/prog", the file descriptor returned is as if your code said open("/usr/tmp/prog", 0). This is of interest to debuggers who need symbol tables, but you are interested in reading from the running process. So read from fd, not nfd. > q3: (curiosity) What is the reason to require the process to be stopped > before doing DFCOPENT or read() from it? This limits the usefulness > of the operations provided. This is due to a sticky implementation issue - I'll try to explain briefly. Our kernel works on multiprocessors as well as uniprocessors. Therefore shared data must be protected from simultaneous access, which implies a lock of some kind. If an arbitrary process was to examine the address space of another processor, the structures that hold this information would have to be locked. Well, in 99.9% of the time, this data is private to the process, and need not be locked. It needs only to be protected when dbx (which is the primary user of /debug) is debugging the process. By stopping a process, we don't need locks since you know that the data structures aren't going to be mucked with. In addition, it is arguably desirable that the process be stopped when being examined by the debugger. In the future, we may be relaxing this restriction by only locking when modifying the virtual address space - a relatively infrequent event. In the meantime, your code above is correct - stop the process and restart it. > Thanks for any info > rayan Hope this helps. Jeff Doughty
rayan@CS.TORONTO.EDU (Rayan Zachariassen) (08/01/89)
Jeff, thanks for your response. It did indeed guide me the right way. My confusion about DFCOPENT was due to its description in the manual page: DFCOPENT If the given argument is zero, return the file descriptor corresponding to the text region of the process. If the given argument is non-zero, interpret this value as a virtual address of the process. Return the file descriptor corresponding to the region containing this address. This call may be used to locate symbol tables of the process. This gave me the impression that opening /debug/<pid> would give only the text segment of the process and one would have to use DFCOPENT with a virtual address to get the stack segment, for example. I got even more confused by seeing strange things in the top page of the stack segment as read from a /debug/<pid> fd... once it was a symbol table, another time it was the contents of the root crontab (which I had edited shortly before). I get the feeling that page isn't being cleared when mapped into the process address space (if so, you will want to fix it, if not, I remain confused). For people listening in, the reason I was wondering about this was to be able to print the hosts people are logged in from in a who listing. We have this (suid, of course) in /local/bin/who. I really hate having to stop login shells like this. #include <stdio.h> #include <sys/types.h> #include <utmp.h> #include <sys/schedctl.h> #include <core.h> #include <sys/fs/dbfcntl.h> #include <sys/signal.h> main(argc, argv) int argc; char *argv[]; { int i; char *at, *host; struct utmp *utp; extern struct utmp *getutent(); extern char *ctime(), *gethost(); /* if this isn't a normal everyday vanilla 'who', punt to the SGI one */ if (argc > 1) { (void) execv("/bin/who", argv); (void) execv("/usr/bin/who", argv); fprintf(stderr, "%s: can't find normal 'who', giving up.\n", argv[0]); exit(1); } /* setup for critical region in gethost() */ /* ignore all signals */ for (i = 1; i < MAXSIG; ++i) (void) signal(i, SIG_IGN); /* up my priority so this runs fast... */ (void) schedctl(NDPRI, 0, NDPHIMAX); while ((utp = getutent()) != NULL) { if (utp->ut_type != USER_PROCESS) continue; at = ctime(&utp->ut_time); at[16] = '\0'; printf("%-8.8s %-8.8s%s", utp->ut_name, utp->ut_line, at+4); if ((host = gethost(utp->ut_pid)) != NULL) printf(" (%s)", host); putchar('\n'); } exit(0); } char * gethost(pid) int pid; { register char *cp; int fd, n, i; static char buf[0x3000]; extern char *strchr(); (void) sprintf(buf, "/debug/%d", pid); if ((fd = open(buf, 0)) < 0) { perror("open"); return NULL; } /* critical section start */ if (fcntl(fd, DFCSTOP, 0) < 0) { perror("fcntl 1"); return NULL; } if (lseek(fd, (long)stackbas(sizeof buf), 0) < 0) { perror("lseek"); return NULL; } if ((n = read(fd, buf, sizeof buf)) < 0) { perror("read"); return NULL; } i = CLEARNOSIG; if (fcntl(fd, DFCRUN, &i) < 0) { perror("fcntl 3"); return NULL; } /* critical section end */ cp = buf + n; while (--cp > buf) { if (*cp == 'R' && *(cp+1) == 'E' && strncmp(cp, "REMOTEHOST=",11) == 0) return strchr(cp, '=') + 1; } return NULL; }