jwp@larry.sal.wisc.edu (Jeffrey W Percival) (05/02/89)
I received several replies to my question in comp.unix.wizards about the "not enough core" message. All pointed out that my swap partition is probably too small for the amount of work being done. One reply mentioned the "swap" entry in fstab, and sure enough, we don't have such an entry. The Question: is the :sw: entry in fstab truly needed, or is the swap info compiled into the kernel? -- Jeff Percival (jwp@larry.sal.wisc.edu)
karish@forel.stanford.edu (Chuck Karish) (05/03/89)
In article <171@larry.sal.wisc.edu> jwp@larry.sal.wisc.edu (Jeffrey W Percival) wrote: [ "not enough core" to run small utilities ] >One reply mentioned the "swap" entry in fstab, and sure enough, >we don't have such an entry. >The Question: is the :sw: entry in fstab truly needed, or is ^^^^^ >the swap info compiled into the kernel? Let's not get overly metaphysical here. When we put the lines into the file, the problem went away. Chuck Karish {decwrl,hpda}!mindcrf!karish (415) 493-7277 karish@forel.stanford.edu
chris@mimsy.UUCP (Chris Torek) (05/03/89)
In article <171@larry.sal.wisc.edu> jwp@larry.sal.wisc.edu (Jeffrey W Percival) writes: [BSDish system] >The Question: is the :sw: entry in fstab truly needed, or is >the swap info compiled into the kernel? Both. The kernel has compiled into it the list of devices it *might ever* swap on. Initially, it swaps/pages/allocates_backing_store_from only the first; the others are added with the vswapon() system call, made by /etc/swapon, which is normally run as swapon -a from /etc/rc. With -a, swapon reads fstab to find :sw: entries and calls vswapon() with the devices found in this way. -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7163) Domain: chris@mimsy.umd.edu Path: uunet!mimsy!chris
guy@auspex.auspex.com (Guy Harris) (05/03/89)
>The Question: is the :sw: entry in fstab truly needed, Yes. >or is the swap info compiled into the kernel? Yes. (Probably.) The problem is that in 4.xBSD's swapping code, as I remember, you have to both 1) tell the kernel that it is to permit swapping on a particular device when you build the kernel (so, in that sense, it's compiled into the kernel, although you should be able to reconfigure this without kernel source) and 2) tell the kernel that it is to *use* a particular swap area when the system comes up multi-user, which is done by putting a ":sw:" entry in "fstab" and running "/etc/swapon" in one of the "/etc/rc*" files ("/etc/swapon" does a "swapon" call to turn swapping on for a particular partition). The only exception is the *first* swapping area, which is built into the kernel and automatically turned on when you boot. I think SunOS 4.0 has lifted restriction 1), which is an improvement.
bph@buengc.BU.EDU (Blair P. Houghton) (05/04/89)
In article <17245@mimsy.UUCP> chris@mimsy.UUCP (Chris Torek) writes: > >The kernel has compiled into it the list of devices it *might ever* >swap on. Initially, it swaps/pages/allocates_backing_store_from only >the first; the others are added with the vswapon() system call, made by >/etc/swapon,[...] I ran into the insufficient-memory bug, too. My sysadmin (the fella sitting to my left) says that the swap space is protected in such a way that an overlarge set of /usr/adm files won't affect it. I was attributing the problem to the fact that we've allowed /usr/adm/acct to hit 17meg. He says they're unrelated. I don't buy it. --Blair
chris@mimsy.UUCP (Chris Torek) (05/04/89)
In article <2760@buengc.BU.EDU> bph@buengc.BU.EDU (Blair P. Houghton) writes: >My sysadmin (the fella sitting to my left) says that the swap space is >protected in such a way that an overlarge set of /usr/adm files won't >affect it. I was attributing the problem to the fact that we've allowed >/usr/adm/acct to hit 17meg. He says they're unrelated. I don't buy it. He is correct. Swap space is allocated only as backing store for processes. It is completely unaffected by file sizes (although file sizes may affect the amount of memory allocated by programs reading or writing those files). Under some systems (SunOS 4.x and Mach), you can swap on ordinary files, so if you run out of VM, you can make a big file and add it to the backing store pool. -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7163) Domain: chris@mimsy.umd.edu Path: uunet!mimsy!chris
guy@auspex.auspex.com (Guy Harris) (05/04/89)
>My sysadmin (the fella sitting to my left) says that the swap space is >protected in such a way that an overlarge set of /usr/adm files won't >affect it. I was attributing the problem to the fact that we've allowed >/usr/adm/acct to hit 17meg. He says they're unrelated. I don't buy it. Uh, well, it depends on what flavor of UNIX you're running, but if your swap space is on a disk partition of its own, he's right. If your system swaps to/from plain files on the same partition that contains "/usr/adm", they could be related. Most UNIXes tend, when swapping to a local disk, to swap to a "raw" partition, since that's what the original AT&T version did. Apollo's isn't derived from AT&T's version at that level, and I think it pages/swaps to/from plain files; I think the same may be true of Mach. Diskless machines running SunOS 4.x can page/swap to an NFS file; others may have modified AT&T-derived UNIXes to page/swap to local or over-the-wire files as well.
bph@buengc.BU.EDU (Blair P. Houghton) (05/05/89)
In article <17287@mimsy.UUCP> chris@mimsy.UUCP (Chris Torek) writes: >In article <2760@buengc.BU.EDU> bph@buengc.BU.EDU (Blair P. Houghton) writes: >>My sysadmin (the fella sitting to my left) says that the swap space is >>protected in such a way that an overlarge set of /usr/adm files won't >>affect it. I was attributing the problem to the fact that we've allowed >>/usr/adm/acct to hit 17meg. He says they're unrelated. I don't buy it. > >He is correct.[...] I'll buy it. It's been pointed out to me that excessive use of X applications can fill up the swap space itself toot sweet, causing the problem I get. He may be correct, but the dozen xterms he runs from .login say he's being a memory-hog. --Blair "All I got is this one li'l window for rn, and my login window, and that spare, and that one with the SPICE runs, and that xsplot with the SPICE graphs, and the two artistically arranged xclocks, and the bitmap of Einstein, glaring at me all the time to imagine more and know less, but THAT's ALL... I swear... :-)"