g_vfs_done and error 11 (EDEADLK)
Anyone know why FreeBSD would give g_vfs_done errors with EDEADLK? Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65968111616, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65968373760, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65968635904, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65968865280, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65969127424, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65969389568, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65969651712, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65969881088, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65970143232, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65970405376, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65970667520, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65970896896, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65971159040, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65971421184, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65971683328, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65971912704, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[READ(offset=81416159232, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():gpt/enc.eli[WRITE(offset=65848246272, length=32768)]error = 11 Aug 22 15:31:14 drum kernel: g_vfs_done():ufs/var[WRITE(offset=1275428864, length=16384)]error = 11 The gpt/enc.eli is a GELI partition. ufs/var is not. If I go to the offset in ufs/var, it turns out to point to the syslog entry corresponding to the previous line. What's really strange is that none of the files on either disk appear to be corrupted. smartctl shows no errors on either drive. dding the entire drive produces no errors. The GELI drive is exported via samba. The errors are triggered when accessing the samba share from a PC. This is not reliably repeatable. My question is: What does error = 11 mean in this context? FreeBSD drum.msfc.nasa.gov 9.2-PRERELEASE FreeBSD 9.2-PRERELEASE #0 r254654: Thu Aug 22 09:13:25 CDT 2013 j...@drum.msfc.nasa.gov:/usr/obj/usr/src/sys/BORODIN9 i386 -- J. Porter Clark jpc2...@inbox.com ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: Status of Chromium port...
On Wed, May 15, 2013 at 11:32:31AM +0300, Volodymyr Kostyrko wrote: 14.05.2013 23:48, Peter Harrison: Hello list! Does anyone know the status of the Chromium port? It's stuck at v25 with multiple vulnerabilities. Updated versions have been available for a while, but haven't been brought into ports. I've emailed the maintainer but not had a response. Anyone know better? I'm building v27 from port now. Looks like many things have changed since v25 - new dependencies, the build flows differently. Seems to be a major update. Indeed, seems a real mess now. I told it not to use pulseaudio, it wants to install it anyway, along with gdbm and accessibility/speech-dispatcher. WTF? Might want to hold off until some of this gets fixed... -- J. Porter Clark j...@porterclark.com ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Samba 3.6.4 winbindd
Has anyone out there gotten winbindd from Samba 3.6.anything to work on FreeBSD 9.0-STABLE? It starts up with no obvious problems--although with Samba's usual cryptic error messages it's hard for me to tell--and then just sits there doing nothing. Wbinfo commands time out and pam_winbind.so doesn't work. When I run /usr/local/etc/rc.d/samba stop, it hangs waiting for winbindd to die, and I have to kill -KILL winbindd. I'm using security = domain. Other settings, logs, etc. available upon request. I've had to drop back to Samba 3.5.14_1 to get most things to work, but I really need to go to 3.6.* for the NTLMv2 support. -- J. Porter Clark j...@porterclark.com ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: How to label a GELI device
On Tue, Jan 25, 2011 at 10:45:52AM +0200, Nikos Vassiliadis wrote: J. Porter Clark wrote: I have an encrypted partition, /dev/da0s1d. I can use geli attach da0s1d and obtain a device /dev/da0s1d.eli, which is a UFS filesystem. All that works just fine. I'd like to label /dev/da0s1d so that I don't have to refer to the exact drive number, etc., which might change if I reboot with a USB stick in the system or whatever. But glabel puts the label in the last sector, which is where GELI stores metadata. You don't have to worry about this. geli uses the last sector for its metadata and creates a device with one sector less to its clients. The original device is 2048 sectors, the device geli provides is 2047 sectors: moby# diskinfo /dev/md0 /dev/md0.eli /dev/md0512 1048576 20480 0 /dev/md0.eli512 1048064 20470 0 There is no way for the internal GEOM to mess with the external's metadata. That's fine, but I want to label the external /dev/md0, not the internal /dev/md0.eli. What I eventually want to do is to geli attach the device using a name that doesn't depend on drive numbering. -- J. Porter Clark j...@porterclark.com ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: How to label a GELI device
On Tue, Jan 25, 2011 at 03:29:37PM +0100, Rolf Nielsen wrote: Correct me if I'm wrong anyone. You need to first label da0s1d e.g. like so glabel label data da0s1d then geli init the labeled device e.g. like so geli init -l 256 -s 4096 label/data Unfortunately, this step overwrites the label. If I try to repeat the glabel command, then the geli metadata is overwritten. That will give you a device node called /dev/label/data.eli, that you can newfs and mount. Unfortunately, since you already encrypted da0s1d, you may have to back it up, and restore the data after you've redone it. I had this problem a few years ago, and I had to back up and restore, but perhaps it's been made simpler now? Though I doubt it. I think that this is the problem. -- J. Porter Clark j...@porterclark.com ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: How to label a GELI device
On Tue, Jan 25, 2011 at 07:28:25PM +0100, Rolf Nielsen wrote: X-Spam-Level: 2011-01-25 19:13, J. Porter Clark skrev: On Tue, Jan 25, 2011 at 03:29:37PM +0100, Rolf Nielsen wrote: Correct me if I'm wrong anyone. You need to first label da0s1d e.g. like so glabel label data da0s1d then geli init the labeled device e.g. like so geli init -l 256 -s 4096 label/data Unfortunately, this step overwrites the label. It does not. I just tested it with a file backed md device, and hexdumped it after each step (creating the file, mdconfig it, label the md device and encrypting it). After the first two steps, I got just zeros, after labeling it, I got the last sector containing the label, and after encrypting it, I got the second last sector (i.e. the last sector of the labeled device) containing the eli data and the last secor still containing the label. If it does overwrite the label, you most likely specified the da0s1d to the geli init command. You need to specify label/data (replace data with the name you choose). Ah! That is, in fact, exactly what I did. I didn't realize that the glabeled device was actually shrunk by 1 sector. Thanks! -- J. Porter Clark j...@porterclark.com ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
How to label a GELI device
I have an encrypted partition, /dev/da0s1d. I can use geli attach da0s1d and obtain a device /dev/da0s1d.eli, which is a UFS filesystem. All that works just fine. I'd like to label /dev/da0s1d so that I don't have to refer to the exact drive number, etc., which might change if I reboot with a USB stick in the system or whatever. But glabel puts the label in the last sector, which is where GELI stores metadata. So, how do I make this work? -- J. Porter Clark j...@porterclark.com ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: Booting multiple choice, and pause to read bootup info
On Tue, Jun 29, 2010 at 04:00:21AM -0500, Robert Bonomi wrote: 1. I'd like to be able to expand the list of choices in the boot menu (the menu with single user mode, safe mode, etc.) to include booting in any of several different environments, e.g., home wired, home wireless, work wired, work wireless. Hacking the FORTH code isn't entirely out of the question, but before I even try it, I need to know how I could tell the system to switch among different rc.conf files (if that's even possible) from the loader. Offhand, I don't see a mechanism for doing so. Cleverer ideas welcome. There's no 'built in' mechanism. There's no easy way. Closest thing -to- an 'easy way' is to set an environment variable _very_early_ in the boot process, and then use it to 'conditionalize' (how -that- for an ugly word? :) the setting of various stuff in rc.conf e.g.: case $USER_ENV in home) USE_LDAP=no ;; work) USE_LDAP=yes ;; esac I wasn't aware that setting an environment variable inside the loader would propagate into the rc.conf environment. Is this so? 2. Usually, when the system boots, there are several lines showing the kernel and various modules loading, possibly with diagnostics. Is there a way to pause after that stage, so that those lines can be read? Or is there any way to retrieve them after the system has booted? I havven't tried it on FBSD, in a long time, but most PC BSDs will pause the boot screen if you hit [CTL-S], or the PAUSE key. Alternatively, does dmesg(8), used 'reasonably soon' after booting, give you what you want? Note: a typical installation will have syslogd putting _most_ of those messsages in the system log file, too. Y'all are way off base here: it's not the lines from the kernel itself booting, it's the lines *before* that, where the loader is loading the kernel and various modules. Occasionally, I see error messages here, but they vanish pretty quickly on my machines, too fast to be caught reliably with CTL-S, SCROLL LOCK, etc. I could set up a serial console, but it seems like a lot of work just to see these messages. -- J. Porter Clark j...@porterclark.com ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: Booting multiple choice, and pause to read bootup info
On Tue, Jun 29, 2010 at 12:09:20PM -0500, Robert Bonomi wrote: _NOT_ directly. But, you can use the loader(8) directive 'init_script' to specify a script that runs 'before anything else'. Either have your menu item execute that word, to set an 'environment' specified by the file, or have a separate menu run _from_ that script. The idea is 'find a hook', then do whatever it takes to use the hook you found. :) Indeed! Seems I had missed the init_script feature; that looks like the easiest solution. Thanks! -- J. Porter Clark j...@porterclark.com ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Booting multiple choice, and pause to read bootup info
1. I'd like to be able to expand the list of choices in the boot menu (the menu with single user mode, safe mode, etc.) to include booting in any of several different environments, e.g., home wired, home wireless, work wired, work wireless. Hacking the FORTH code isn't entirely out of the question, but before I even try it, I need to know how I could tell the system to switch among different rc.conf files (if that's even possible) from the loader. Offhand, I don't see a mechanism for doing so. Cleverer ideas welcome. 2. Usually, when the system boots, there are several lines showing the kernel and various modules loading, possibly with diagnostics. Is there a way to pause after that stage, so that those lines can be read? Or is there any way to retrieve them after the system has booted? -- J. Porter Clark j...@porterclark.com ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: Nanobsd on a CD-ROM
On Thu, Jun 12, 2008 at 06:54:45PM +0200, Pietro Cerutti wrote: J. Porter Clark wrote: | On Thu, Jun 12, 2008 at 04:40:25PM +0200, Pietro Cerutti wrote: | J. Porter Clark wrote: | | Is it possible to build a CD-ROM with a bootable NanoBSD on it? | | If so, how? | | Yes, Section 2.2 of | | http://www.freebsd.org/doc/en/articles/nanobsd/index.html | | Well, no, because I can't do this: | | # dd if=_.disk.full of=/dev/acd0 bs=64k | | If I do this: | | % cdrecord -v -immed driveropts=burnfree dev=1,0,0 -data _.disk.full Try with burncd | burncd -f /dev/acd0 data _.disk.full fixate No joy. Produces the same disk that cdrecord does, and boots from hard disk instead. I think that this is the problem: The BIOS knows how to boot from a CD if and only if that CD is an El Torito bootable image. That is, the first sector of the CD is NOT a Master Boot Record. That's just a hypothesis based on the observation that all of the successfully bootable CDs I have appear to be in El Torito format. I don't have a way to make such an image without using mkisofs. Figuring out how to pack the nanobsd image in such a way that mkisofs can make an El Torito bootable CD from it sounds difficult, offhand. Anybody know how to do this sort of thing? Is it even possible? -- J. Porter Clark [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Nanobsd on a CD-ROM
On Thu, Jun 12, 2008 at 04:40:25PM +0200, Pietro Cerutti wrote: J. Porter Clark wrote: | Is it possible to build a CD-ROM with a bootable NanoBSD on it? | If so, how? Yes, Section 2.2 of http://www.freebsd.org/doc/en/articles/nanobsd/index.html Well, no, because I can't do this: # dd if=_.disk.full of=/dev/acd0 bs=64k If I do this: % cdrecord -v -immed driveropts=burnfree dev=1,0,0 -data _.disk.full then try to boot the machine from the CD-R, I get a - cursor for a second or two, then it switches over to the hard disks to boot from. Okay, it's possible I messed up the compilation options or something. Is the cdrecord command above the correct procedure to use to build a bootable nanobsd CD-R? -- J. Porter Clark [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Nanobsd on a CD-ROM
Is it possible to build a CD-ROM with a bootable NanoBSD on it? If so, how? -- J. Porter Clark [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Making mergemaster skip certain files
Is there any way to keep certain files out of the reach of mergemaster? I understand the need for carefully merging the old and the new, but I really shouldn't ever have to for files like these: /etc/aliases /etc/hosts /etc/hosts.allow /etc/manpath.config ... and many others. Mergemaster has so many options that I'm fairly certain that there must be some way to do this. -- J. Porter Clark [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Okay, who broke sysutils/eject?
eject from /usr/ports/sysutils/eject used to work fine. Now, on a 4.9-STABLE system: % eject -v acd0c eject: trying device /dev/acd0cc eject: trying device /dev/acd0c eject: /dev/ad0s3a mounted on / eject: /dev/ad0s3f mounted on /tmp eject: /dev/ad0s3g mounted on /usr eject: /dev/ad0s3e mounted on /var eject: unmounting /var eject: /var: Device busy And on a different system: % eject -v acd0c eject: trying device /dev/acd0cc eject: trying device /dev/acd0c eject: /dev/da0s1a mounted on / eject: /dev/da0s1f mounted on /tmp eject: /dev/da0s1g mounted on /usr eject: /dev/da0s1e mounted on /var eject: /dev/da1s1e mounted on /home eject: unmounting /home eject: /home: Device busy In neither case is the CD (or anything else) ejected. What's going on? -- J. Porter Clark [EMAIL PROTECTED] ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Nvidia and Dell Latitude's cover switch
I've been using x11/nvidia-driver from ports on a Dell Latitude C840 laptop with 4.8-STABLE. I have it set to do 1600x1200. The problem is that whenever I close the cover while it's running X, and then reopen the cover, the display is screwed up. Specifically, it looks like the left-hand half of the screen has been stretched to cover the full screen, and the right-hand half is gone. When this happens, I can recover by doing Ctrl-Alt-F1 and then Ctrl-Alt-F9. Or I can cycle through video modes with Ctrl-Alt-KP+. But sheesh, it's annoying as all get out. Below are the relevant parts of my XF86Config file. Any idea how to fix this problem? BTW, I had no problems like this with the VESA driver. Section Device Option HWcursor True Option ShadowFB True Option UseFBDev False Option FlatPanel True Option FPDither True Identifier Card0 Driver nvidia VendorName NVIDIA BoardName GeForce4 440 Go Option NvAgp 1 EndSection Section Screen Identifier Screen0 Device Card0 Monitor Monitor0 DefaultDepth24 SubSection Display Depth 24 Modes 1600x1200 1280x1024 1024x768 800x600 640x480 EndSubSection EndSection -- James Porter Clark[EMAIL PROTECTED] NASA/MSFC Computers and Data Systems Group (ED13) ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]