cdl@mplvax.UUCP (Carl Lowenstein) (06/04/84)
Operating System Design - the XINU Approach. Douglas Comer, Prentice-Hall, Inc. 1984 Comer's book is a neat approach to operating systems but there are lots of little things I can detect that bother me. This lowers my confidence about the possibility of little things that I can't detect because of the level of my ignorance. First, and nit-pickiest: the computer under discussion throughout the book is called the LSI 11/02. The official DEC name is LSI-11/2 if you are talking about the CPU board itself, and PDP-11/03 if it's in a box. Next, my contribution to the book's index -- all the header files "*.h" should be in there. A visual grep has produced the following: bufpool.h 234 frame.h 264 proc.h 54 conf.h 142 iblk.h 310 q.h 44 dir.h 318 io.h 116 sem.h 84 disk.h 285 kernel.h 72 sleep.h 130 dlc.h 213 mark.h 232 slu.h 168 dtc.h 284 mem.h 105 tty.h 161 file.h 320 ports.h 242 xebec.h 282 Also note that the index entries for conf.c and conf.h are secondary references, and should be 152 and 142, respectively. Now for the rest of the nits, in page order: pp. 24-25 Highest memory address is 0175777, should be 0157777 (ok in sizemem.s, p. 191) p. 25 Start of I/O page is 0176000, should be 0160000 p. 30 Transmitter CSR bit 7 has 0 - 1 sense reversed (ok in kprintf.c, p. 358) p. 34 Fig. 2.11 is missing one byte (ok in xebec.h, p. 282) p. 60 _ctxsw has stack manipulation which will fail if the routine is interruptible. Can do the same thing cleanly with no more instructions or processor time. Also, code for loading new process registers doesn't use auto-increment, and thus takes more space and time than it really needs. p. 136 Three classic examples of "off by one" in a loop counter. "dec, bpl" doesn't branch until decremented past 0, because 0 is positive. "dec, bne" is the appropriate code. Or else change the meaning of each counter. p. 199 Ex. 13.1 Rewrite the Xinu downloader . . . I'd like to try, but it isn't in the book. p. 284 & p. 387 "Digital Technology" was "Data Technology" when I bought my disk controllers. It's Data Technology on p. 34. Now for some real bottom line questions: Does someone have exact instructions for getting from Prentice-Hall the tape that goes with this book? (name, address, order number, price, etc.) Would someone who already has the tape comment on the following: Is it necessary to have a VAX as the central support computer, or would a larger PDP-11 do the job? On the other hand, does the support software requre compatibility mode on the VAX, or would a micro-VAX do the job? Probably this discussion (if any) should be moved to net.misc, unless someone thinks there should be a net.usoft.XINU. Where does one discuss other operating systems, anyhow? Net.XINU-wizards :-) Carl Lowenstein sdcsvax!mplvax!cdl Marine Physical Lab., U.C. San Diego