Re: C++ code in a kernel module?
On Mon, Sep 08, 2003 at 11:12:59PM -0400, Alexander Kabaev wrote: On Mon, 8 Sep 2003 23:02:33 -0400 Matthew Emmerton [EMAIL PROTECTED] wrote: I've been silently following this thread, and unless I missed something, has anyone asked John why he wants/needs to use C++ in the kernel? Tools, not policy :) True but if John explains more of what he is trying to do, maybe someone can suggest either a way to avoid C++ or some insights as to how he can get his C++ modules to work. Peter ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Perl 5.8.1-RC4 release notes mention FreeBSD malloc upto 200 times slower than perl malloc
In the Perl 5.8.1-RC4 release notes posted here
http://dev.perl.org/perl5/news/2003/perl-5.8.1-RC4.html
it mentions about the Platform specific problems seen by the Perl
development team.
--
FreeBSD: malloc dilemma
The choice of malloc (the C-level memory management interface) when
building Perl is problematic in FreeBSD.
Using FreeBSD's system malloc for Perl was found to be very slow: in
some cases that was 200 times slower than using the Perl malloc. One
such case is file input: for example
# slurping the whole compressed Perl source code into $a
if (open F,perl-5.8.1.tar.gz) { local $/; $a=F }
is about 200-250 times slower with the system malloc than with the Perl
malloc.
One could use Perl's malloc (Configure -Dusemymalloc), but that was
found to cause random core dumps in FreeBSD with multithreaded programs.
No such problems were found in other platforms, however.
A decision was made to stick with the system malloc, regardless of the
performance problems.
--
Maybe the test case can assist FreeBSD kernel hackers to make malloc
faster.
Regards, Yusuf
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-hackers
To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Serial ATA 4.8 Installation -help
It seems florian mettetal wrote: Greetings, I have built a brand new system founded on an Asus P4PE motherboard, and using the FastTrack raid function I have turned my two Maxtor 80GB Serial ATA (now reffered to as SATA) hard drives into a Mirror raid. Proceeding to start the installation of FreeBSD 4.8-stable I selected Standard Install, Press ok at the pop up, then at the following popup I am confronted with FreeBSD telling me that No disks found! Please verify that your disk controller is being properly probed at boot time. Has anyone else had this issue? Is this problem fixed in FreeBSD 5.X . What can I do to get this to work in FreeBSD 4.8? 4.x doesn't support the Promise SATA controllers, you need at least 5.1 for that. The -current ATA driver might be backported later this year, nothing definite yet. -Søren ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: linprocfs issues
Steven Hartland [EMAIL PROTECTED] writes: Also it seems that linprocfs relies on procfs being mounted however I cant find any docs that mention this and see no reason why it should be the case. The problem I see is that I get errors from linux apps about /proc/pid/file not existing. When trussing them they are never trying to open file I suspect it is happening when the access either cmdline or exe most likely exe. linprocfs uses parts of the procfs code; for instance, linprocfs's pid/exe and procfs's pid/file are backed by the same code. In the case of pid/cmdline, which exists with the same name in both FreeBSD and Linux, linprocfs relies on procfs's cmdline to show through rather than duplicate it. DES -- Dag-Erling Smørgrav - [EMAIL PROTECTED] ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: OpenBFS (was Re: C++ code in a kernel module?)
On Mon, 8 Sep 2003, Pedro F. Giffuni wrote: Hi; Attached is a good reasons why someone my want to use C++ in the kernel. Sorry, I don't see anything here except this is all we know how to do. But, I'm a curmudgeon. :) Doug -- This .signature sanitized for your protection ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Booting a machine over the network without pxe.
On Tue, 2 Sep 2003, Josef Karthauser wrote: Here is How i have made this task: http://www.freebsd.org/cgi/query-pr.cgi?pr=54383 I guess that is exactly what you want. I have posted a PR, and maybe the FreeBSD devel team think about it. :) []'s Leal Date: Tue, 2 Sep 2003 21:54:18 +0100 From: Josef Karthauser [EMAIL PROTECTED] To: [EMAIL PROTECTED] Subject: Booting a machine over the network without pxe. Does anyone have any experience of booting a machine over the network, like pxeboot, but without running PXE on a network card. I imagine that it should be possible to load pxeboot at the boot: prompt and have everything just work. I could really do with booting my laptop into -stable, where it's only got -current installed. I do however have a -stable server on site with plenty of disk space. It would be really cool to remote boot of that via NFS mounts, etc. Joe -- Josef Karthauser ([EMAIL PROTECTED]) http://www.josef-k.net/ FreeBSD (cvs meisterk admin and hacker) http://www.uk.FreeBSD.org/ Physics Particle Theory (student) http://www.pact.cpes.sussex.ac.uk/ An eclectic mix of fact and theory. = [EMAIL PROTECTED] SDF Public Access UNIX System - http://sdf.lonestar.org ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Perl 5.8.1-RC4 release notes mention FreeBSD malloc upto 200 times slower than perl malloc
On Tue, Sep 09, 2003 at 03:00:17PM +0800, Yusuf Goolamabbas wrote: Maybe the test case can assist FreeBSD kernel hackers to make malloc faster. This is a FAQ. For any algorithm one can come up with a workload that makes it perform badly. The trick is making it perform well for common workloads, which FreeBSD's malloc does. Kris pgp0.pgp Description: PGP signature
Re: C++ code in a kernel module?
On Tuesday, Sep 9, 2003, at 00:29 America/Denver, Peter Jeremy wrote: On Mon, Sep 08, 2003 at 11:12:59PM -0400, Alexander Kabaev wrote: On Mon, 8 Sep 2003 23:02:33 -0400 Matthew Emmerton [EMAIL PROTECTED] wrote: I've been silently following this thread, and unless I missed something, has anyone asked John why he wants/needs to use C++ in the kernel? Tools, not policy :) True but if John explains more of what he is trying to do, maybe someone can suggest either a way to avoid C++ or some insights as to how he can get his C++ modules to work. Peter simple, I have preexisting C++ code which we are currently using in userland and wish to push down into the kernel. It would be ideal to keep the source bases the same without a rewrite to C. Admitting of course the possibility of having to modify to be compatible with both use modes. At present I am attempting to see what we can and cannot do in the kernel with C++ John G ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: C++ code in a kernel module?
On Tuesday, September 9, 2003, at 08:58 AM, John Giacomoni wrote: On Tuesday, Sep 9, 2003, at 00:29 America/Denver, Peter Jeremy wrote: On Mon, Sep 08, 2003 at 11:12:59PM -0400, Alexander Kabaev wrote: On Mon, 8 Sep 2003 23:02:33 -0400 Matthew Emmerton [EMAIL PROTECTED] wrote: [snip] simple, I have preexisting C++ code which we are currently using in userland and wish to push down into the kernel. It would be ideal to keep the source bases the same without a rewrite to C. Admitting of course the possibility of having to modify to be compatible with both use modes. At present I am attempting to see what we can and cannot do in the kernel with C++ FWIW, Darwin (the underpinnings for Mac OS X) uses C++ for its device drivers. This is done by hewing to a model roughly that of Embedded C++, compiling statically, and having a separate library that differs from libstdc++ in significant ways. Getting this to work well was non-trivial, but it does work (sadly :-}). If you are trying to get user-mode code to work in the kernel, you are in for an enjoyable year... You can check Apple's Darwin site for available doc (http://developer.apple.com/darwin). The code is available under Apple's open source license (APSL 2.0). Regards, Justin -- /~\ The ASCII Justin C. Walker, Curmudgeon-at-Large \ / Ribbon Campaign X Help cure HTML Email / \ ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: C++ code in a kernel module?
I don't know about __gxx_personaility_v0, but your best bet is to look at the .o's and find where it is referenced. Then back track it to what function, then to what construct and go from there. The __gxx_personality_v0 stuff is actually related to distinguishing between C++ and Java exceptions. At least that's what I remember from my encounters with the same problem. I haven't been able to retrieve the documentation that explains it, though. It seems to come up when you try to link C++ code with the C linking convention. -- Fred Gilham [EMAIL PROTECTED] Lisp has jokingly been called the most intelligent way to misuse a computer. I think that description is a great compliment because it transmits the full flavor of liberation: it has assisted a number of our most gifted fellow humans in thinking previously impossible thoughts. E. Dijkstra ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Perl 5.8.1-RC4 release notes mention FreeBSD malloc upto 200 times slower than perl malloc
On Tue, Sep 09, 2003 a.d., Yusuf Goolamabbas wrote:
# slurping the whole compressed Perl source code into $a
if (open F,perl-5.8.1.tar.gz) { local $/; $a=F }
I fortunately fixed that in perl 5.8.1, so when slurping regular files
the buffer is no longer grown by little steps with realloc.
Regards,
Adi
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-hackers
To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: OpenBFS (was Re: C++ code in a kernel module?)
--- Doug Barton [EMAIL PROTECTED] wrote: On Mon, 8 Sep 2003, Pedro F. Giffuni wrote: Hi; Attached is a good reasons why someone my want to use C++ in the kernel. Sorry, I don't see anything here except this is all we know how to do. But, I'm a curmudgeon. :) To be good in C++ you have to be good in C first (at least in theory) and programmers that feel more modern tend to use OO techniques for the new stuff. I have to admit I don't find much of an advantage in C++, but it doesn't make much sense to rewrite existing code or change your programming style to maintain everything in C. Just my $0.02, Pedro. __ Do you Yahoo!? Yahoo! SiteBuilder - Free, easy-to-use web site design software http://sitebuilder.yahoo.com ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: C++ code in a kernel module?
In message: [EMAIL PROTECTED] Matthew Emmerton [EMAIL PROTECTED] writes: : I've been silently following this thread, and unless I missed something, has : anyone asked John why he wants/needs to use C++ in the kernel? I refrained from posting the name of a good neurologist in Boulder that he could go see to get his head checked. Warner ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
NFS mounts and jails
I've successfully been exploring using NFS mounts within a jail. It's clear to me that a host box running multiple jails needs to expose that NFS mount to each jail. As far as I know, that means that NFS partition has to be mounted (redundantly?) for each jail, even in (in my case), they'll all be access in the the same way (eg. read-only). This seems to be that you'll quickly run into some effective upper limit as to how many NFS mounts a FreeBSD machine can handle (I'm using 4.7-R for my current experiments.) So, my question(s): - is my assertion correct; that is, do I need to redundantly mount the same NFS partition for each jail for each jail to access it? - if not, is there a mechanism for jails sharing a read-only NFS mountpoint? Thanks for any advice... -- Brian 'you Bastard' Reichert[EMAIL PROTECTED] 37 Crystal Ave. #303Daytime number: (603) 434-6842 Derry NH 03038-1713 USA BSD admin/developer at large ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: C++ code in a kernel module?
I can't believe nobody has yet suggested that FreeBSD switch to IOKit... On Sep 9, 2003, at 12:38 PM, M. Warner Losh wrote: In message: [EMAIL PROTECTED] Matthew Emmerton [EMAIL PROTECTED] writes: : I've been silently following this thread, and unless I missed something, has : anyone asked John why he wants/needs to use C++ in the kernel? I refrained from posting the name of a good neurologist in Boulder that he could go see to get his head checked. Warner ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED] -- Jordan K. Hubbard Engineering Manager, BSD technology group Apple Computer ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: C++ code in a kernel module?
Apart from -fno-rtti and -fno-exceptions you will probably need to use also -ffreestanding. Excerpt from gcc man page: -ffreestanding Assert that compilation takes place in a freestanding environment. This implies -fno-builtin. A freestanding environment is one in which the standard library may not exist, and program startup may not necessarily be at main. The most obvious example is an OS kernel. This is equivalent to -fno-hosted. Vaclav Haisman ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Basic C++ fBSD 5.x kernel module
I have succeeded in getting a basic C++ module to function inside the FreeBSD 5.1 kernel. I figured I'd show the small example I have cooked up which uses basic C++ constructs such a new/delete and virtual inheritance. As suggested I have turned off exceptions have been turned off along with RTTI. the source and Makefile can be found at: http://www.cs.colorado.edu/~jgiacomo/fbsd_c++module Things to note are: a) I rolled my own .cc.o extension to the Makefile removed -std=c99 added -fno-exceptions -fno-rtti -fno-builtin -nostdinc++ (-fno-builtin should be implied by -ffreestanding but does not seem to be) b) __unused does not seem to work and needed to be #undef'ed * Any thoughts? c) Constructs of the form static void const * const seem to be silently ignored by the compiler. Possibly being 'inlined' in some fashion? * Any thoughts? part is 'c' is problem as various macros, such as MALLOC_DEFINE and DEV_MODULE, use that construct for initialization. thoughts? comments? suggestions? John G ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: NFS mounts and jails
As far as I know, that means that NFS partition has to be mounted (redundantly?) for each jail, even in (in my case), they'll all be access in the the same way (eg. read-only). Yup, that's right. - is my assertion correct; that is, do I need to redundantly mount the same NFS partition for each jail for each jail to access it? As far as I know, yes that's the case. If you're mounting the same system however you may want to look into mount_null. Yes, there are cache issues, and these have been discussed multiple times (see the archives), but I've used null mounts successfully in read-only mode for several years now without problems. Cheers, Nate ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Serial ATA 4.8 Installation -help
On Tue, 2003-09-09 at 02:54, Soren Schmidt wrote: It seems florian mettetal wrote: Greetings, I have built a brand new system founded on an Asus P4PE motherboard, and using the FastTrack raid function I have turned my two Maxtor 80GB Serial ATA (now reffered to as SATA) hard drives into a Mirror raid. Proceeding to start the installation of FreeBSD 4.8-stable I selected Standard Install, Press ok at the pop up, then at the following popup I am confronted with FreeBSD telling me that No disks found! Please verify that your disk controller is being properly probed at boot time. Has anyone else had this issue? Is this problem fixed in FreeBSD 5.X . What can I do to get this to work in FreeBSD 4.8? 4.x doesn't support the Promise SATA controllers, you need at least 5.1 for that. The -current ATA driver might be backported later this year, nothing definite yet. -Sren Hmm... I have never setup a mirror Raid before, especially never in FreeBSD 5.1 , I have always been using 4.8 =/ Now, I am at the Select Drives (Second screen) which menas that 5.1 does see my promise SATA raid controller. Now... I have 3 selections to use for drives, ad4, ad6, ar0 First, I have 2 drives, and one is just a mirror of the other. my presumption is to singularly use ar0, but... I really do not know. And as this will someday hopefully be my production server... I wanna set it up right. Any clues anyone? ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: NFS mounts and jails
On Wed, Sep 10, 2003 at 01:08:50AM +, Nielsen wrote: - is my assertion correct; that is, do I need to redundantly mount the same NFS partition for each jail for each jail to access it? As far as I know, yes that's the case. If you're mounting the same system however you may want to look into mount_null. Yes, there are cache issues, and these have been discussed multiple times (see the archives), but I've used null mounts successfully in read-only mode for several years now without problems. Thanks for the feedback, let me explore that option... Cheers, Nate -- Brian 'you Bastard' Reichert[EMAIL PROTECTED] 37 Crystal Ave. #303Daytime number: (603) 434-6842 Derry NH 03038-1713 USA BSD admin/developer at large ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
FYI: NetBSD mmap Improvements
taken from wine HQ: http://www.winehq.com/?issue=186#NetBSD mmap Improvements On NetBSD (upcoming 1.6.2, and 1.7/2.0-current), there is a new extension flag MAP_TRYFIXED that essentially simulates current Linux mmap behavior: try the mmap() hint first, without clobbering mapped pages, even if the hint falls within traditionally protected malloc heap space. If the fixed mapping fails, the block is still mapped at a relocated address, as if mmap were called with no flags set. With this patch, mmapping PE files on NetBSD becomes an order of magnitude faster, as the vfork()-and-mincore() silly walk is avoided altogether. I've implemented the patch as forward-looking, allowing other platforms to add MAP_TRYFIXED to gain the same benefit. (This mmap flag name does not appear to be used in any divergent fashion on any other platform, per my research when picking the flag's name.) __ Do you Yahoo!? Yahoo! SiteBuilder - Free, easy-to-use web site design software http://sitebuilder.yahoo.com ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Booting a machine over the network without pxe.
omestre wrote this message on Tue, Sep 09, 2003 at 14:22 +: On Tue, 2 Sep 2003, Josef Karthauser wrote: Here is How i have made this task: http://www.freebsd.org/cgi/query-pr.cgi?pr=54383 I guess that is exactly what you want. I have posted a PR, and maybe the FreeBSD devel team think about it. :) boot.netif.hostname already exists as kern.hostname. Why create yet another knob for the same value? Why don't you make it parse nfsroot.* from vfs.root.mountfrom? That way you can specify it as nfs:host:/path. -- John-Mark Gurney Voice: +1 415 225 5579 All that I will do, has been done, All that I have, has not. ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hackers To unsubscribe, send any mail to [EMAIL PROTECTED]
