Re: Daily kernel panics on 4.6-RC system
Here's my /etc/make.conf: CFLAGS= -O -pipe COPTFLAGS= -O -pipe NO_KERNELCLEAN= true NO_MAKEDEV= true MODULES_WITH_WORLD= true Maybe worth noting is that the buildworld buildkernel targets were built with 'make -j4' on a dual CPU machine. I might comment out the COPTFLAGS entry and omit the -j4 option to make, but then -j4 -O should be safe shouldn't they? I might try swapping the RAM with another box to see if that helps. Cheers, Mark - Original Message - From: Ian Dowse [EMAIL PROTECTED] To: Mark Ibell [EMAIL PROTECTED] Cc: [EMAIL PROTECTED] Sent: Thursday, May 30, 2002 8:52 AM Subject: Re: Daily kernel panics on 4.6-RC system In message 008b01c206b8$981dbcb0$166e10ac@MARKI, Mark Ibell writes: We have had trouble with our Dell PowerEdge 1400 server since day one and it has now got to the point where it kernel panics once or twice a day. The 1400 currently runs 4.6-RC but we have had a similar problem with 4.3-RELEASE 4.5-RELEASE, although the problem has definitely got a lot worse since upgrading to 4.6-RC. Perhaps someone can make sense of the following crash dump info. Fatal trap 12: page fault while in kernel mode fault virtual address = 0x50 fault code = supervisor read, page not present #6 0xc024100b in tcp_output (tp=0xde1ef7c0) at /usr/src/sys/netinet/tcp_output.c:942 This crash shouldn't happen; the code is trying to read tp-rcv_nxt in the line if (win 0 SEQ_GT(tp-rcv_nxt+win, tp-rcv_adv)) but somehow the non-NULL `tp' argument has become NULL during the execution of the function. In the assembly output here, `tp' is kept in one register throughout the whole tcp_output() function, so it must be getting corrupted by one of the functions called by tcp_output(). Are you using any unusual compiler optimisations in /etc/make.conf? There are known issues with using -O2 or greater that would certainly explain these crashes. It would also be worth trying different hardware to determine if this is a software or hardware issue. Ian To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: IPC Messages
Fabio Yasusi Yamamoto wrote: Hi How can i increase msg_qbytes 2048 ? ( without recompile the kernel ) thx! This isn't really a -stable question, it should have been posted to questions. But sysctl might do it, man 8 sysctl to start. I haven o access to a freebsd machine at the moment otherwise i'd be of more help. --Larry To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: Hardware RAID vs vinum
On Wed, 2002-05-29 at 07:11, Scott Long wrote: On Wed, May 29, 2002 at 01:42:11PM +0930, Greg 'groggy' Lehey wrote: I'm sure that there are good reasons for small stripe sizes. I don't believe that efficiency is one of them. I think your problem is that you're looking at single-request latency. That's usually not an issue, though it may impact copy performance. As I said, though, that's not usually where you would use RAID-5. The real issue is not so much latency as throughput. I hope I've been able to make it more understandable. It is obvious that I have tread onto religious territory here, and as such there is probably little I can say to convince you of my points. I think that we must then agree to disagree =-) Au contraire. There is no religion involved. Greg presented hard facts. You can either refute his facts by presenting your own hard facts or else concede the point. If you cannot refute the facts, then you, by implication, concede the argument. /Joe To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: Hardware RAID vs vinum
On Wednesday, 29 May 2002 at 8:11:41 -0600, Scott Long wrote: On Wed, May 29, 2002 at 01:42:11PM +0930, Greg 'groggy' Lehey wrote: I'm sure that there are good reasons for small stripe sizes. I don't believe that efficiency is one of them. I think your problem is that you're looking at single-request latency. That's usually not an issue, though it may impact copy performance. As I said, though, that's not usually where you would use RAID-5. The real issue is not so much latency as throughput. I hope I've been able to make it more understandable. Greg, It is obvious that I have tread onto religious territory here, Not for me. I've given you some very detailed arguments. I also have measurements which back them up. and as such there is probably little I can say to convince you of my points. Well, you could start by trying. I think that we must then agree to disagree =-) I'd much rather see some technical arguments from you. Greg -- See complete headers for address and phone numbers To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: Hardware RAID vs vinum
I am not going to take a stand on performance other then to say to both parties: prove your points. I personally believe that features and management characteristics are better comparison points between software and hardware RAID systems. Performance is and always will be a moving target. But I will interject one point of interest, and that is the cost factor. As a friend of mine pointed out to me a few days ago, the cost of RAID hardware has dropped precipitously over the last 5 years. It is cheap enough now that one might as well just use it on multi-disk servers. Hopefully this will result in more and better drive support in FreeBSD, I sometimes get the feeling that we've dropped behind in that area. -Matt Matthew Dillon [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: VMware 3.1.1 for Windows and FreeBSD 4.6-RC2
FYI: Same stories is applied to 4.6-RC3. I've tried to install FreeBSD 4.6-RC2 on VMware 3.1.1 for Windows box. However, * kern.flp boots fine, loader started. * loader reads mfsroot.flp contents cleanly. * kernel starts. Display USERCONFIG menu. Skip it. * kernel detects devices. * sysinstall starts, then Probing devices, please wait (this can take a while)... dialog is shown. * sysinstall seems to be stalled. so I cannot install. Installing 4.5-RELEASE works fine with the same VM configuration. What am I missing? OK, other users on [EMAIL PROTECTED] (a japanese list) found that sysinstall doesn't stall if VM configuration has no ATAPI CD-ROM. When I change my VM config to have SCSI CD-ROM instead, sysinstall doesn't stall, happily shows the first installation menu. I know that Mar/30/2002's 4-stable goes fine, since I've checked that it works before committing src/usr.sbin/sysinstall/devices.c rev 1.139. However, Apr/15/2002's 4-stable does not. Anybody knos what's the difference? *** P.S.: Speaking about atomic_cmpset_int mentioned by iedowse, my problem is not related to this, since 4-stable's GENERIC kernel already defines I386_CPU. -- - Makoto `MAR' Matsushita To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: VMware 3.1.1 for Windows and FreeBSD 4.6-RC2
matusita I know that Mar/30/2002's 4-stable goes fine, since I've matusita checked that it works before committing matusita src/usr.sbin/sysinstall/devices.c rev 1.139. I try to reconfirm this, but Mar/30/2002's boot floppies doesn't work also, sorry. Maybe I didn't check on VMware... -- - Makoto `MAR' Matsushita To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: Hardware RAID vs vinum
On Wednesday, 29 May 2002 at 17:16:36 -0700, Matthew Dillon wrote: I am not going to take a stand on performance other then to say to both parties: prove your points. Well, I thought I had done just that. If you have any issues with my claims, you're welcome to raise them. I personally believe that features and management characteristics are better comparison points between software and hardware RAID systems. Performance is and always will be a moving target. What I'm talking about is how to configure a system which gives you the choice of stripe sizes. That's orthogonal to the other issues. Greg -- See complete headers for address and phone numbers To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
gprof pthreads, crash (null pointer)
Can anyone tell me if gprof (ie compiling with -pg) is supposed to work with pthreads? I'm finding I get a null pointer from pthread_create(). Actually, the program counter becomes a null pointer, as in: #0 0x in ?? () #1 0x08052cd3 in osiThread::Start( ... ) I'm using gcc30 from the ports, specifically gcc30-3.0.4. on FreeBSD 4.6-RC #2: Wed May 22 19:30:53 EDT 2002 $ gcc30 -v Reading specs from /usr/local/lib/gcc-lib/i386-portbld-freebsd5.0/3.0.4/specs Configured with: ./..//gcc-3.0.4/configure --disable-nls --with-gnu-as --with-gnu-ld --with-gxx-include-dir=/usr/local/lib/gcc-lib/i386-portbld-freebsd5.0/3.0.4/ include/g++ --disable-shared --prefix=/usr/local i386-portbld-freebsd5.0 Thread model: posix gcc version 3.0.4 I'm compiling everything with -pg. To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: VMware 3.1.1 for Windows and FreeBSD 4.6-RC2
Ok, I've investigated a little, and found that sysinstall is stalled at line 342 of src/release/sysinstall/devices.c rev 1.117.2.16. In function deviceTry(), sysinstall try to open(2) /dev/acd0c: at first, there is no device file, it should be failed. sysinstall next try to make a device file /dev/acd0c with 117:0 by mknod(2). After that, sysinstall try to open(2) /dev/acd0c again. It successes, and return with file descriptor. However, when sysinstall close(2) the file descriptor last opened, sysinstall stalled. It seems that close(2) syscall doesn't return. Note that this behavior is occured both CD-ROM drive has a disc or not. It would be a problem of 4-stable kernel, and VMware. Anybody runs recent (read: Apr/2002 or later) 4-stable on VMware? Your virtual ATAPI CD-ROM works fine for you? Or, is there something wrong with following code? main() { int fd; fd = open(/dev/acd0c, O_RDONLY); close(fd); /* stalls here */ } -- - Makoto `MAR' Matsushita To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: Server won't boot after recompile the kernel with ipfw support
On Wednesday, May 29, 2002, at 12:46 PM, Ian wrote: Have a look at /etc/rc.firewall and see if one of the standard named configurations it supports is right for you. If so, set it in firewall_type in rc.conf. Otherwise use firewall_type=/etc/ipfw.conf and put your own ruleset into that file (which won't get clobbered on upgrades). This thread is questionably still on -stable, but wouldn't it make more sense to set firewall_script to your ruleset file instead of firewall_type? ie, I use: firewall_enable=YES firewall_script=/etc/rc.firewall.local Setting firewall_type to a file name will just ensure that no rules are added at all, it won't match any cases in /etc/rc.firewall. Bryan To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Error in make buildworld (cvs 5-29 @ 11:30 CDT)
Was greeted by the below (bottom) errors during make buildworld earlier. Tried rm'ing the /usr/src/gnu tree and re-running cvsup, but got the same error. Make is run with no flags at the command line. Relevant flags from make.conf: CPUTYPE=i686 BDECFLAGS= -W -Wall -ansi -pedantic -Wbad-function-cast -Wcast-align \ -Wcast-qual -Wchar-subscripts -Winline \ -Wmissing-prototypes -Wnested-externs -Wpointer-arith \ -Wredundant-decls -Wshadow -Wstrict-prototypes -Wwrite-strings Otherwise, it's just /etc/defaults/make.conf. Suggestions/comments welcome. Thanks, Rob cc -O -pipe -march=pentiumpro -I/usr/src/gnu/usr.bin/groff/src/preproc/pic/../../../../../../contrib/groff/src/preproc/pic -I. -DHAVE_STDLIB_H=1 -DHAVE_UNISTD_H=1 -DHAVE_DIRENT_H=1 -DHAVE_LIMITS_H=1 -DHAVE_STRING_H=1 -DHAVE_STRINGS_H=1 -DHAVE_MATH_H=1 -DRET_TYPE_SRAND_IS_VOID=1 -DHAVE_SYS_NERR=1 -DHAVE_SYS_ERRLIST=1 -DHAVE_CC_LIMITS_H=1 -DRETSIGTYPE=void -DHAVE_STRUCT_EXCEPTION=1 -DHAVE_GETPAGESIZE=1 -DHAVE_MMAP=1 -DHAVE_FMOD=1 -DHAVE_STRTOL=1 -DHAVE_GETCWD=1 -DHAVE_STRERROR=1 -DHAVE_PUTENV=1 -DHAVE_RENAME=1 -DHAVE_MKSTEMP=1 -DHAVE_STRCASECMP=1 -DHAVE_STRNCASECMP=1 -DHAVE_STRSEP=1 -DHAVE_STRDUP=1 -DSYS_SIGLIST_DECLARED=1 -I/usr/src/gnu/usr.bin/groff/src/preproc/pic/../../../../../../contrib/groff/src/include -I/usr/src/gnu/usr.bin/groff/src/preproc/pic/../../../src/include -static -o pic lex.o main.o object.o pic.o common.o troff.o tex.o /usr/obj/usr/src/i386/usr/src/gnu/usr.bin/groff/src/preproc/pic/../../../src/libs/libgroff/libgroff.a -lm /usr/libexec/elf/ld: k_cos.o: invalid string offset 27 = 0 for section `' /usr/libexec/elf/ld: k_cos.o: invalid string offset 33 = 0 for section `' /usr/libexec/elf/ld: k_cos.o: invalid string offset 43 = 0 for section `' /usr/lib/libm.a(s_sin.o): In function `__generic_sin': s_sin.o(.text+0xa3): undefined reference to `__kernel_cos' s_sin.o(.text+0xdf): undefined reference to `__kernel_cos' /usr/lib/libm.a(s_cos.o): In function `__generic_cos': s_cos.o(.text+0x7b): undefined reference to `__kernel_cos' s_cos.o(.text+0xb7): undefined reference to `__kernel_cos' *** Error code 1 Stop in /usr/src/gnu/usr.bin/groff/src/preproc/pic. *** Error code 1 Stop in /usr/src/gnu/usr.bin/groff/src/preproc. *** Error code 1 Stop in /usr/src/gnu/usr.bin/groff/src. *** Error code 1 Stop in /usr/src/gnu/usr.bin/groff. *** Error code 1 Stop in /usr/src. *** Error code 1 Stop in /usr/src. *** Error code 1 Stop in /usr/src. To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: Server won't boot after recompile the kernel with ipfw support
Setting firewall_type to a file name will just ensure that no rules are added at all, it won't match any cases in /etc/rc.firewall. Scanning rc.network quickly, it looks like you're correct for recent -STABLE. In which case the following comments in rc.firewall should be reaped, surely?: # Define the firewall type in /etc/rc.conf. Valid values are: # open - will allow anyone in # client - will try to protect just this machine # simple - will try to protect a whole network # closed - totally disables IP services except via lo0 interface # UNKNOWN - disables the loading of firewall rules. # filename - will load the rules in the given filename (full path required) # msg45894/pgp0.pgp Description: PGP signature
Re: Hardware RAID vs vinum
But I will interject one point of interest, and that is the cost factor. As a friend of mine pointed out to me a few days ago, the cost of RAID hardware has dropped precipitously over the last 5 years. It is cheap enough now that one might as well just use it on multi-disk servers. Hopefully this will result in more and better drive support in FreeBSD, I sometimes get the feeling that we've dropped behind in that area. Mind if I squeak in one little comment on the state of hardware RAID card support...? I wish I could back it up with come code, but I'm just a lowly sysadmin :) (maybe one day, in one way or another) While as far as I can tell RAID card support is quite stable (knock on wood, thanks guys and girls), there's a noticable absence of RAID array control status reporting software for many types of RAID cards, including my own (NetRAID 3i, amrd driver). For instance, for my particular type of RAID controller I cannot tell if the array has degraded without walking up to the server and seeing amber warning lights :(Good thing I'm not using these cards in a colo :) Even a simple syslog message saying Houston, we have a problem would be invaluable.. :) Of course I'm definitely not expecting the world, just though I'd point this out :) By the way I'm using either RELENG_4_5 or -stable on my boxes... Oh, and a question: can software RAID mirrors (via vinum) be booted from? (and still booted from if one drive goes down?) Thanks very much, Eric Parusel To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message