From: Digestifier To: Linux-Development@senator-bedfellow.mit.edu Reply-To: Linux-Development@senator-bedfellow.mit.edu Date: Tue, 4 Oct 94 11:13:08 EDT Subject: Linux-Development Digest #263 Linux-Development Digest #263, Volume #2 Tue, 4 Oct 94 11:13:08 EDT Contents: Does Linux require an IRQ for SCSI (George) Odd floppy sector size? (Dale Shuttleworth) Re: umount problem! (iafilius@et.tudelft.nl) Re: ext2fs vs. Berkeley FFS (Ian McCloghrie) ext2fs vs. Berkeley FFS (Hugh Strong) Re: Ip Adrress probs (Jim Williams) Re: ext2fs vs. Berkeley FFS (Hugh Strong) Re: p5 code optimization for gcc (Shaune Beattie) Re: What GUI to write for? (Kevin K. Lewis) Re: Linux/Warp2 HPFS improper shutdown flag set.. (Ralf G. R. Bergs) Re: Korn Shell '93 Now Available from AT&T (Peter Mutsaers) Re: Is there a memory allocation debugging tool for Linux? (Mitchum DSouza) Re: Does Linux require an IRQ for SCSI (Jeff Kesselman) Re: Improving SLIP latency under Linux (Christopher Michael Joslyn) Re: Multiprocessing Pentium Systems (Huw Davies) ---------------------------------------------------------------------------- From: ghharrac@ouray.Denver.Colorado.EDU (George) Subject: Does Linux require an IRQ for SCSI Date: 2 Oct 1994 03:00:56 GMT Does linux require that SCSI devices use an IRQ. The seagate SCSI driver finds my adapter and drive, but times out. The interrupt selector is disabled on my card! Under DOS the adapter uses no DMA channels, no I/O ports, and no interrupts. All operations are memory mapped I/O. The system uses 8K of memory space at address CA00h. I can set the IRQ to 3|5, Chance the Wait state to 0, and modify the Memory range. Any Suggestions, is an IRQ required ? ghharrac@ouray.denver.colorado.edu ------------------------------ From: dale@giskard.demon.co.uk (Dale Shuttleworth) Subject: Odd floppy sector size? Reply-To: dale@giskard.demon.co.uk Date: Sun, 2 Oct 1994 02:37:21 GMT Hi, I've got a bit of a problem. I'm trying to read floppies written with 256 byte sectors. Looking at the floppy code for 1.1.45, there doesn't seem to be a coherent strategy for handling different sector sizes. In some places it is assumed to be 512 bytes, in others it is replaced with a #define, which may be either 128 or 512 bytes. I know that the code is changing quite a lot with other features, but is there any reason why the facility to handle arbitary (within the constraints of the FDC) was not included. I have to admit that I know absolutely nothing about the PC FDC so I could of course be trying to achieve the impossible. If it is feasible, does anybody see any reason why I shouldn't do it? I can see that I might run the risk of breaking several IOCTLs if I try to tack on storage for a sector size indicator to the existing data structures. Any comments, cries of don't do it, pleas to do it, indications that it can't be done, etc, are most welcome. Dale. -- ****************************************************************************** * Dale Shuttleworth * * Email: dale@giskard.demon.co.uk * ****************************************************************************** ------------------------------ From: iafilius@et.tudelft.nl Subject: Re: umount problem! Date: 4 Oct 94 01:13:43 +0100 In article <36pnu7$lp2@vixen.cso.uiuc.edu>, roth@ux4.cso.uiuc.edu (Mark D. Roth) writes: > I just encountered a weird problem with umount. I am using the umount > that comes with the binary distrib of util-linux-1.10 under linux > 1.1.51. When I mount a floppy and move files to it (the filesystem > was ext), and then umount it > > Anyway, I dunno what's causing this. You are not allone in the dark, I have similar problems after upgrading the kernel. And I have not any idee what causing it. So people help us please. Arjan =================================== / * _____ / / / / / / \ / / / / / / / \/ / / / / / / /\ /____ / / / /_____/ / \ ================================== Arjan Filius Email: iafilius@et.tudelft.nl Tell them : Get Linux Don't even think of BUYING msdos and windows. windows is for the masses Linux is for the smarter ones! ------------------------------ From: ianm@qualcomm.com (Ian McCloghrie) Subject: Re: ext2fs vs. Berkeley FFS Date: 3 Oct 1994 10:47:50 -0700 hstrong@eng1.uconn.edu (Hugh Strong) writes: >is only one aspect of the equation. It may be reasonable to >compare ext2fs to FFS on this basis, but FFS dates from about >(I think) 4.1 or 4.2 BSD. I think the issue today is flexibility >as much as it is performance. 4.3, in fact. (may even have been 4.3 Tahoe). The FFS was one of the big things 4.3 gave over 4.2. -- Ian McCloghrie work: ianm@qualcomm.com home: ian@egbt.org ____ GCS d-- H s+:+ !g p? au a- w+ v- C++$ UL++++ US++$ P+>++ \bi/ L+++ 3 E+ N++ K--- W--- M-- V-- -po+ Y+ t+ 5+++ jx R G''' \/ tv- b+++ D- B-- e- u* h- f+ r n- y* The above represents my personal opinions and not necessarily those of my employer, Qualcomm Inc. ------------------------------ From: hstrong@eng1.uconn.edu (Hugh Strong) Subject: ext2fs vs. Berkeley FFS Date: 2 Oct 1994 08:27:18 GMT Just wondering - How does the performance of Linux ext2fs compare with that of the Berkeley Fast File System (FFS) found in {386,Free,Net}BSD and other BSD variants? A number of posts to the 386BSD groups have recently sneered at ext2fs, presumably because of the considerations FFS makes for drive geometry and rotational parameters, which seem to be absent in the ext2fs source I've examined. Does anyone have any concrete performance statistics to back/refute a these claims? Is anyone workrking on FFS for Linux? - Hugh Strong ------------------------------ From: jim@iceworld.org (Jim Williams) Subject: Re: Ip Adrress probs Date: 2 Oct 1994 00:37:13 GMT : You need to do a few things. First, you need to enable IP forwarding : in your kernel if you're using any kernel after about 1.1.13. This is : one of the "make config" options. : Second, you need to set up some routing. On the gateway, you need to : set up a default route to the .100 address. You also need to set up a : route for your internal network. : On each machine (on your internal network), the default gw will be the : gateway machine's address (whatever that is). : Finally, you need to be able to tell your provider about your internal : network. This is most easily done with a static route at the provider, : which will identify the gateway machine as the next hop on the way to : your internal network. This is the easiest way, but not necessarily : the best. The best way is probably to run "routed" or "gated" and send : RIP updates to your provider from the gateway machine. However, this : takes some more work to get set up. : I would recommend an investment in O'Reilly's "TCP/IP Network : Administration", available at most (good) computer stores or technical : bookstores. It explains how to do all of this. You can also reach : O'Reilly via WWW or ftp to ora.com. : Good luck! You are a life saver. Thanks for the info. I will try all of the above. I have gated and it is not running but will look over the docs and set that up. I also have purchesed the TCP/IP Book above. It was a good investment. I will let ya know how things work out. Thanks agian. Jim jim@iceworld.org ------------------------------ From: hstrong@eng1.uconn.edu (Hugh Strong) Subject: Re: ext2fs vs. Berkeley FFS Date: 3 Oct 1994 01:41:22 GMT Rob Janssen (rob@pe1chl.ampr.org) wrote: : In <36lqt6$t80@babyblue.cs.yale.edu> hstrong@eng1.uconn.edu (Hugh Strong) writes: : >Just wondering - : >How does the performance of Linux ext2fs compare with that of : >the Berkeley Fast File System (FFS) found in {386,Free,Net}BSD and other : >BSD variants? A number of posts to the 386BSD groups have recently : >sneered at ext2fs, presumably because of the considerations FFS makes for : >drive geometry and rotational parameters, which seem to be absent in the : >ext2fs source I've examined. Does anyone have any concrete performance : >statistics to back/refute a these claims? Is anyone workrking on : >FFS for Linux? : Well, the Linux community sneers at BSD for doing synchronous inode : updates, which you won't find in ext2fs :-) : Those can really kill performance when you are manipulating a lot of : files, like in a news system. : Conclusion: The filesystems have different characteristics, and the : performance will vary with the operations you are doing. When you want : to know what will perform best in your situation, try them both. : But please don't try to setup a war between the two camps! I think it's important to keep in perspective that performance is only one aspect of the equation. It may be reasonable to compare ext2fs to FFS on this basis, but FFS dates from about (I think) 4.1 or 4.2 BSD. I think the issue today is flexibility as much as it is performance. Ext2fs supports a limited set of extended attributes (append, immutable, secure delete, uneraseable, synchronous) and the sources indicate that ACLs will be available in the future. But these are all kernel space extensions. You can't associate some arbitrary piece of data (that stays with the file when copied) with the file without writing it in the data stream somewhere. By contrast, HPFS allows up to 64K of extended attributes per file. NTFS goes the full distance by allowing multiple arbitrary streams per file, although there doesn't seem to be any documentation on how to exploit this in the Win32 API manual. I know this doesn't sound very UNIXy, but how difficult would it be to implement a filesystem with some sort of arbitrary extended attributes like those in NTFS or the more limited ones in HPFS? -- Hugh Strong hstrong@ug1.eng1.uconn.edu ------------------------------ From: sdgb1@cus.cam.ac.uk (Shaune Beattie) Subject: Re: p5 code optimization for gcc Date: 4 Oct 1994 02:59:28 GMT ChiWei Che (werche@saba.rutgers.edu) wrote: : Hi all: : A faithful Linux user here. Recently, I got a Pentium machine, of course the : thing is dedicated to Linux. Yesterday, I am upgrading the gcc 2.5.6 to : intel optimized version of gcc 2.5.8. I use that to recompile the kernel, : the system is running faster before (although I didn't run any benchmark, and : I don't know any benchmark in Linux, but you definitely can feel it). To all : who are using the P5 CPU, try to replace the old one with pentium code, it's : better. BTW, my system configuration: Hmm, I downloaded this one but had nothing but problems,. I could get it to compile, then compile itself and i can use the new -mpentium flag... *but* if i use -O4 to compile anything more than a 'hello world' program, then gcc crashes with an internal error. If i use the -O3 flag it will compile, but the reulting code will be very unstable. And as the performance gain on just specifying the -mpentium and staying at -O2, seemed to not be that great, I changed back to the 486 version. (I benchmarked it by running povray on a large trace.. maybe not the best test in the world but the difference was negible) btw i got the gcc-i2.5.8p off tsx in /pub/linux/ALPHA/something (or something like that) presume that is the only version.? Shaune ------------------------------ From: lewikk@grasshopper.aud.alcatel.com (Kevin K. Lewis) Crossposted-To: comp.windows.x.intrinsics,gnu.misc.discuss Subject: Re: What GUI to write for? Date: 04 Oct 1994 13:26:52 GMT In article jwshin@nitride.EECS.Berkeley.EDU (Jinwoo Shin) writes: Yes, Motif is very ugly IMHO and I don't think Athena is overpriced at all :) For me xview hits the spot. Code freely available, and its API is not too bad and GUI is aesthetically pleasing in my eyes. Good example being workman. There is currently a project underway (according to the Linux Projects Map) to clone Motif. The result is supposed to be freely available. I'm not sure what the license will look like, though. I think combining this with FWF code is an ideal choice. -- Kevin K. Lewis | My opinions may be unreasonable lewikk@aud.alcatel.com | but such is the voice of inspiration ------------------------------ From: rabe@pool.informatik.rwth-aachen.de (Ralf G. R. Bergs) Crossposted-To: comp.os.os2.beta Subject: Re: Linux/Warp2 HPFS improper shutdown flag set.. Date: Mon, 03 Oct 1994 17:24:38 +0200 =====BEGIN PGP SIGNED MESSAGE===== In article , pinaar@netftp.austin.ibm.com (Arthur J Pina) wrote: > Does anyone know of a fix for this? > > Someone reported earlier that Linux sees the OS/2 Warp2 HPFS partitions > as having the "improper shutdown" flag set -- well I just experienced this > problem. -- fortunately I had my Linux "doom" code in an OS/2 V2.11 partition, > so I could still access it, but, this leaves me with less accessable HPFS > space then I would like. in linux/fs/hpfs/hpfs.h change the line unsigned dirty; /* 0 clean, 1 "improperly stopped" */ to unsigned dirty : 1; /* 0 clean, 1 "improperly stopped" */ unsigned incognita : 31; /* more flags */ Hope that helps, Ralf - -- Ralf G. R. Bergs, Hueckeswagener Str. 42, 51647 Gummersbach, FR Germany |Team phone/fax: +49 2261 21968, E-Mail: rabe@pool.informatik.rwth-aachen.de |OS/2 =====BEGIN PGP SIGNATURE===== Version: 2.6ui iQCVAgUBLpAwhWCy3Wi6qSJ1AQFBAwP/cT1ItKzqWtuwWoudDrniMQyUxttk7Q4t 2N6dkNklqL+uvkOLZdj81Qxrtj/TMTMdFVOSa9/xPjzebidDQpc2rCXPTWcbyWym D3w+dV/ri8febW+dbh3Q/9qEHA67M2mHFQ8V1AbeNn1rU9zKOQetQ1V2JAC9gaa8 JUHG+eHBVwA= =FsJJ =====END PGP SIGNATURE===== -- Ralf G. R. Bergs, Aachen University of Technology EE (comp. eng.) student snail: H"uckeswagener Str. 42, D-51647 Gummersbach, Fed. Rep. of Germany phone/fax: +49 2261 21968 E-Mail: rabe@pool.informatik.rwth-aachen.de Click here. ------------------------------ From: plm@atcmp.nl (Peter Mutsaers) Subject: Re: Korn Shell '93 Now Available from AT&T Date: Tue, 4 Oct 1994 13:00:58 GMT >> On 29 Sep 1994 20:01:15 GMT, mcnalley@metnet.geog.pdx.edu (James >> E. McNalley) said: JEM> No it doesn't. What does ksh do that BASH can't do? Both have their strengths and weaknesses. I would choose bash however. At my work the rest of my colleagues use ksh. bash is clearly superior in interactive use: for example nicer keybindings (you can use arrows with history), easy filename expansion (ksh's is clumsy) and command completion (ksh lacks this). ksh has more complicated programming constructs that one might use in shell-scripts (such as types and arrays). I don't need these however, because I want either that my shell scripts are bourne-shell compatible, or if I have to make something more complicated that needs such extra features I use something more powerful like perl or C. -- Peter Mutsaers | AT Computing bv, P.O. Box 1428, plm@atcmp.nl | 6501 BK Nijmegen, The Netherlands tel. work: +31 (0)80 527248 | tel. home: +31 (0)3405 71093 | "... En..., doet ie het al?" ------------------------------ From: Mitchum.DSouza@mrc-apu.cam.ac.uk (Mitchum DSouza) Subject: Re: Is there a memory allocation debugging tool for Linux? Date: 4 Oct 1994 09:55:53 GMT In article , fox@graphics.cs.nyu.edu (David Fox) writes: |> In article <36mhko$7ua@news.uni-c.dk> hjh@snake8.imsor.dth.dk (Henrik Juul |> Hansen) writes: |> |> ] "debug-malloc" |> ] found at: |> ] oak.oakland.edu:/pub2/unix-c/languages/c/debug-malloc.tar.Z No one mentioned just about the best alloc debugger available for Linux (probably on par with the likes of Purify(TM) IMHO) which is Checker. Check (no pun intended) out sunsite.unc.edu:/pub/Linux/devel/c/Checker-libs-V0.5.tgz Mitch ------------------------------ From: jeffpk@netcom.com (Jeff Kesselman) Subject: Re: Does Linux require an IRQ for SCSI Date: Sun, 2 Oct 1994 05:00:08 GMT In article <36l7p8$ruc@carbon.denver.colorado.edu>, George wrote: >Does linux require that SCSI devices use an IRQ. The seagate >SCSI driver finds my adapter and drive, but times out. The >interrupt selector is disabled on my card! > >Under DOS the adapter uses no DMA channels, no I/O ports, and no >interrupts. All operations are memory mapped I/O. The system uses >8K of memory space at address CA00h. > >I can set the IRQ to 3|5, Chance the Wait state to 0, and modify >the Memory range. > >Any Suggestions, is an IRQ required ? > >ghharrac@ouray.denver.colorado.edu > > Well, my Trantor 130B runs fine with IRQ disabled, but it might be a driver issue. I'm afraid I don't know anything about the segate driver. (I guess that isn't much help, but at least I can tell you its not a Linux requirement...) ------------------------------ From: chrisj@pvi.com (Christopher Michael Joslyn) Subject: Re: Improving SLIP latency under Linux Date: Mon, 3 Oct 1994 19:14:39 GMT To: nickkral@po.EECS.Berkeley.EDU Subject: Re: Improving SLIP latency under Linux Newsgroups: comp.os.linux.development In-Reply-To: <36m9lc$6q0@agate.berkeley.edu> Organization: Visual Numerics, Inc. Cc: Bcc: In article <36m9lc$6q0@agate.berkeley.edu>, Nick Kralevich wrote: > >Is there any way to improve/derease the latency associated with >SLIP under linux? Specifically, when I am ftping a large file, >I frequently get ping times of 6+ seconds. This murders interactive >traffic. I've tried setting my MTU to 256, but it doesn't make >any difference. The main problem is that an ftp packet is much larger than, say, a telnet or ping packet. Because the ftp packet is large, the packets must be broken up into sizes that can be sent over the line (this is your MTU at work), thus the ftp packet takes longer to send. Additionaly, an ftp connection typically is doing things in a batch where it is constantly pushing these large packets onto the queue. Factor all things in and what you get is a long queue filled mostly with ftp packets and a few, notably smaller, interactive packets. >There was a thread a couple of months ago that said the problem >was in the kernel. However, there was never a solution posted. Not really. It is a network problem. Lars Fenneberg (lf@gimli.comlink.de) and I are currently working on a package that would balance the load on a SLIP or PPP dial-up network connection. Unfortunantly, we are just getting started and will be a little bit before it is complete. The implemention at this time simple slices the bandwidth into equal sections so each connection has an equal bandwidth. What we are planning on adding is unequal sharing. Interactive connections would probably have more of the bandwidth since they usually transmit in bursts. The ftp connection would then be slowed during these bursts and then be back up to speed when nothing else is communicating. Of course, this kind of unequal sharing will be tunable at runtime. If you have any further questions/coments/etc., please either contact me (chrisj@boulder.vni.com) or Lars. - Chris -- ---- Purgamentum Init, Exit Purgamentum ------------------------------------ Christopher M. Joslyn | Visual Numerics, Inc. | "A la fin de l'envoi, +1 (303) 581-3269 | 6230 Lookout Road | je touche!" chrisj@boulder.vni.com | Boulder, CO 80301 | My opinions are mine! ------------------------------ Subject: Re: Multiprocessing Pentium Systems From: cchd@lucifer.latrobe.edu.au (Huw Davies) Date: Tue, 4 Oct 1994 03:09:53 GMT David Monro (davidm@syd.dms.CSIRO.AU) wrote: : As an aside, are there any really good MP OSs out there? It's a bit off topic, but OpenVMS is not a bad implementation of a MP OS. If you belive the usually reliable netnews, there was a port of user space VMS to Mach running on 386s too.... -- Huw Davies | Huw.Davies@latrobe.edu.au Computing Services | Phone: +61 3 479 1500 Fax: +61 3 479 1999 La Trobe University | I own an Alfa to keep me poor in a monetary Melbourne Australia | sense, but rich in so many other ways ------------------------------ ** FOR YOUR REFERENCE ** The service address, to which questions about the list itself and requests to be added to or deleted from it should be directed, is: Internet: Linux-Development-Request@NEWS-DIGESTS.MIT.EDU You can send mail to the entire list (and comp.os.linux.development) via: Internet: Linux-Development@NEWS-DIGESTS.MIT.EDU Linux may be obtained via one of these FTP sites: nic.funet.fi pub/OS/Linux tsx-11.mit.edu pub/linux sunsite.unc.edu pub/Linux End of Linux-Development Digest ******************************