corwin@polari.UUCP (Don Glover) (08/28/89)
Well yet another thing that I don't understand about xenix that maybe someone can help me with. I decided to try and tune my system and so I changed some configuration parameters and made a xenix.new. I rebooted the system and specified xenix.new, everything seems to be working fine, but I noticed that the new kernel is 230K smaller than the old one. It is 299K while the old what abot 530K, this worries me alot. Does anyone know what may have happened? I would not expect that the configuration options would have made that much difference I expected to gain some but not 230k, I cut my multi screens from 12 to 8 I specified nbuff at 40 and nhbuf at 256 and set dmaexecl to 0. That was all.
clay@uci.UUCP (News Administrator) (08/28/89)
The size of the XENIX kernel changes radically with the number of multiscreens configured. If the kernel determines the number dynamically, the size is small (or if the number genned is small, natch) If a specific number is specified, there is a lot of static storage for the screens in the kernel image. It doesn't necessarily mean less memory usable in your system, just that the kernel is getting it in a different way. -- Clayton Haapala ...!mmm!dicome!uci!clay Unified Communications Inc. 3001 Metro Drive - Suite 500 "Revenge is better than Christmas" Bloomington, MN 55425 -- Elvira
richard@neabbs.UUCP (RICHARD RONTELTAP) (08/29/89)
[ Xenix kernel from 530K to 230K ] If you haven't got any fancy stuff like X/Windows or TCP/IP linked in your kernel, 230K is about the right size. I wonder how you ever got the 530K kernel... Richard (...!hp4nl!neabbs!richard)
jcw@jwren.UUCP (John C. Wren) (08/29/89)
In article <200959@neabbs.UUCP> richard@neabbs.UUCP (RICHARD RONTELTAP) writes: >[ Xenix kernel from 530K to 230K ] > >If you haven't got any fancy stuff like X/Windows or TCP/IP linked in >your kernel, 230K is about the right size. > >I wonder how you ever got the 530K kernel... One thing I have noticed is that the sysadm shell script for linking a new kernel is not the same as the /usr/sys/conf/sys/link_xenix script. I seem to recall that the sysadm script does not do a strip on the newly generated kernel... Perhaps this could be it? - John C. Wren jcw@jwren
ivar@acc.uu.no (Ivar Hosteng) (08/31/89)
jcw@jwren.UUCP (John C. Wren) writes: >In article <200959@neabbs.UUCP> richard@neabbs.UUCP writes: >>[ Xenix kernel from 530K to 230K ] >> >>If you haven't got any fancy stuff like X/Windows or TCP/IP linked in >>your kernel, 230K is about the right size. >> >>I wonder how you ever got the 530K kernel... >One thing I have noticed is that the sysadm shell script for linking a new >kernel is not the same as the /usr/sys/conf/sys/link_xenix script. I seem >to recall that the sysadm script does not do a strip on the newly generated >kernel... Perhaps this could be it? Stripping the kernel is not wery smart. By doing this you makes it impossible for the ps, wmstat and the w command (and more, I dont remember them all) to find the user & proc structures inside the kernel image. This means they will fail. My Xenix file is 324002 bytes long and I have all multiscreens enabled by default. -- Ivar E. Hosteng, Advanced Computer Consultans, Oslo, Norway Internet: ivar@acc.uu.no UUCP: ...!{uunet,mcvax,ifi}!acc.uu.no!ivar 'Just what do you think you are doing Dave?' -HAL9000
corwin@polari.UUCP (Don Glover) (08/31/89)
In article <200959@neabbs.UUCP>, richard@neabbs.UUCP (RICHARD RONTELTAP) writes: > I wonder how you ever got the 530K kernel... > > Richard > (...!hp4nl!neabbs!richard) Actually, I think I may have figured it out. I rebooted the old kernel and looked closely at the info screen as it came up. I did the same with the new kernel. The major dif was in the fact that on the old kernel the system determined the number of i/o buffs automaticaly, which it did to the tune of 303 of them at 1k each and in the new kernel I had specified only 40 i/o buffs, and so had gained around 260K of free space. I wounder what a true save lower limit on i/o buffs is for 2 user systme or even a 5 user system.
jim@applix.UUCP (Jim Morton) (09/01/89)
> In article <200959@neabbs.UUCP> richard@neabbs (RICHARD RONTELTAP) writes: > One thing I have noticed is that the sysadm shell script for linking a new > kernel is not the same as the /usr/sys/conf/sys/link_xenix script. I seem > to recall that the sysadm script does not do a strip on the newly generated > kernel... Perhaps this could be it? The configure and link_xenix process I believe differs from running a make on the kernel in that the mini assembler in configure puts some data in the DATA section, whereas the make with the Development System puts that data in BSS. This will result in a size difference in /xenix of 100k+. The only real difference though is the amount of on-disk storage the kernel takes up -- if you compare the boot time kernel size numbers for a kernel built each way they come out the same... You can see this happen for yourself by watching the .s files in /usr/sys/conf. (Don't strip the kernel!) -- Jim Morton, APPLiX Inc., Westboro, MA ...uunet!applix!jim jim@applix.com
richard@neabbs.UUCP (RICHARD RONTELTAP) (09/03/89)
> > In article <200959@neabbs.UUCP> richard@neabbs (RICHARD RONTELTAP) writes: > > One thing I have noticed is that the sysadm shell script for linking a new > > kernel is not the same as the /usr/sys/conf/sys/link_xenix script. I seem > > to recall that the sysadm script does not do a strip on the newly generated > > kernel... Perhaps this could be it? This quote about suggesting to strip the kernel is *not* from me (Richard Ronteltap (good name, huh)), but from somone who anwered my message. Stripping the kernel is strictly prohobited. Richard (...!hp4nl!neabbs!richard)