Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×
Bug Debian Operating Systems Software Windows Linux

Strange Ubuntu/Vista Compatibility Bug, Solved 140

Walter Vos writes "Since I've been running Vista and Ubuntu in dual boot with a shared FAT32 partition for my personal folders, I've been seeing some strange compatibility issues between these two operating systems. Somehow Vista locks the folders on the FAT32 partition that are used for folders like Documents, Downloads, etc. A blogpost I wrote gives a detailed description of the problem and a fix for it."
This discussion has been archived. No new comments can be posted.

Strange Ubuntu/Vista Compatibility Bug, Solved

Comments Filter:
  • FAT32 (Score:5, Interesting)

    by Enderandrew ( 866215 ) <enderandrew@NOsPAM.gmail.com> on Thursday August 14, 2008 @09:38PM (#24609519) Homepage Journal

    NTFS-3G works pretty well. I'm not sure FAT32 is really necessary any more.

    • Re:FAT32 (Score:5, Funny)

      by duckInferno ( 1275100 ) on Thursday August 14, 2008 @09:47PM (#24609607) Journal
      Stay away from FATMAN239. It nuked my hard drive.
    • Re: (Score:3, Insightful)

      by 4D6963 ( 933028 )
      ntfs-3g worked pretty well for me, except for I/O intensive applications. aMule with all its I/O on a NTFS partition of VMware with all the virtual machine's file on a NTFS partition as well were pretty slow. Actually I think VMware was so slow that 99% of the CPU was actually taken up by ntfs-3g, meaning VMware was crawling.
      • Re:FAT32 (Score:5, Interesting)

        by MindlessAutomata ( 1282944 ) on Thursday August 14, 2008 @10:44PM (#24610099)

        I kept all my mp3s on an NTFS partition, and it made amarok incredibly slow for searching through files and even listing them when I wanted to expand a tree. It, of course, also was using up a ton of cpu power. Other intensive programs were causing me other problems, mostly more cpu usage quirks.

        NTFS-3g is not perfect and I'd recommend steering clear of relying on NTFS on linux for heavy or day-today usage. I haven't used ubuntu on windows but I can imagine it would give a negative impression due to performance issues. For pulling off the occasion file off another partition, though, it works well.

        When I moved all my mp3s to an ext3 partition, all the problems with amarok went away instantly.

        • Re: (Score:3, Informative)

          by coolsnowmen ( 695297 )

          I kept all my mp3s on an NTFS partition, and it made amarok incredibly slow for searching through files and even listing them when I wanted to expand a tree. It, of course, also was using up a ton of cpu power. Other intensive programs were causing me other problems, mostly more cpu usage quirks.

          I found the default database backend slow, so switching to a better DB could be the solution. Even if your files are on NTFS, try having a postgres DB backend(on your fs of choice) and it should speed up your library searching.

        • Re: Ubuntu and NTFS (Score:4, Informative)

          by szaka ( 1061180 ) on Friday August 15, 2008 @08:10AM (#24613129)
          NTFS-3G changes rapidly and historically Ubuntu included an old, lower performing version of the NTFS-3G driver. However the one in Ubuntu 8.04 should be ok.

          Amarok has a documented performance issue with NTFS-3G: http://ntfs-3g.org/support.html#dd [ntfs-3g.org]

          The NTFS-3G web site has many tips what could be the problem for high CPU usage: http://ntfs-3g.org/support.html#cpu100 [ntfs-3g.org]

          Sometimes NTFS defragmentation makes a magic.

          The focus of the NTFS-3G development is reliability and functionality over performance. The performance optimizations started only recently and the current development versions perform close or sometimes surprisingly even better than ext3.

        • My ancient Pismo Powerbook running OS X seemed like it would be much more useful if I installed NTFS-3G on it so I could use my external (NTFS) hard drive and watch movies and play music, because it certainly has the power to decode the video.

          The only problem was, after I started using it everything really slowed down, it was taking half of my processor just to access file on the hard drive.

          I guess my only real option is to find an old firewire hard drive I can format in HFS.

      • IO is always slow. But there are some tricks to make it faster. But most of the performance tricks lose their advantages with FUSE (e.g. mmap'ed IO or direct IO or raw io, whatever sales call it today). VMware uses bunch of the tricks to actually speed up it's IO and it really works well with normal in-kernel file systems. But not with FUSE.

        • by szaka ( 1061180 )
          But most of the performance tricks lose their advantages with FUSE

          Why? Current results show that a FUSE file system can be even faster than kernel file systems, e.g. ext3:

          Some write speeds from http://www.nabble.com/Re%3A-Porting-Zfs-features-to-ext2-3-p18722897.html [nabble.com]

          tmpfs: 975 MB/sec
          ntfs-3g: 889 MB/sec (note, this FUSE driver is not optimized yet)
          ext3: 675 MB/sec
          • And where are the latencies?

            And how well safety of data are ensured??

            Throughput alone is meaningless to real world.

            Layers, FUSE has to add, send latencies high and reliability down. I wouldn't trust my data to it. Though as nice workaround I gladly use it ;)

            And in case of VMware, raw i/o is actually used as a way to improve latencies and memory consumption by bypassing OS file cache. Guest OSs normally have their own caches anyway.

    • Re:FAT32 (Score:5, Informative)

      by niteice ( 793961 ) <icefragment@gmail.com> on Thursday August 14, 2008 @09:54PM (#24609675) Journal
      There's a couple of ext3 drivers for Windows (one open-source, one not) that also work pretty well, so you can go both ways.
      • Re: (Score:2, Insightful)

        by Anonymous Coward

        As far as I know they're only ext2 drivers. Of coarse, you can usually mount ext3 as ext2 without any issues.

        • Re: (Score:1, Interesting)

          by Anonymous Coward

          As far as I know they're only ext2 drivers. Of coarse, you can usually mount ext3 as ext2 without any issues.

          What keeps people from implementing ext3 support for Windows? The Linux source code is obviously available, so are Windows ext2 drivers reimplementations that aren't using existing code? Or is there some deeper problem?

          • Re:FAT32 (Score:5, Insightful)

            by kiddygrinder ( 605598 ) on Thursday August 14, 2008 @11:01PM (#24610225)
            ext2 works pretty well for ext3 drives so they don't care enough to do it. Anyone who does care about ext3 that much i'd guess probably doesn't care that much about windows.
          • by tepples ( 727027 ) <tepplesNO@SPAMgmail.com> on Thursday August 14, 2008 @11:29PM (#24610437) Homepage Journal

            What keeps people from implementing ext3 support for Windows? The Linux source code is obviously available, so are Windows ext2 drivers reimplementations that aren't using existing code? Or is there some deeper problem?

            For a while, Microsoft once charged roughly $1,000 for the "IFS Kit" used to develop installable file system drivers. To work around this, programs such as "Explore2fs" had to act like WinRAR and 7-Zip, where you don't really mount a partition but you can still drag files in and out. (The price appears to have dropped since then.) For another thing, 64-bit versions of Windows Vista put an annoying "Test Mode" banner in all four corners of the desktop if the user installs a device driver that hasn't been signed by a publisher who pays an annual fee of at least $200 to a commercial certificate authority trusted by Microsoft.

            • Re: (Score:2, Insightful)

              by ozphx ( 1061292 )

              At least $200! Thats almost two developer hours of money!

              Pretty certain you can chuck whatever cert you want in the trusted root store / disable this behaviour.

              • Re: (Score:3, Insightful)

                by tepples ( 727027 )

                At least $200! Thats almost two developer hours of money!

                In what city of what state/province of what country?

                • | At least $200! Thats almost two developer hours of money!

                  In what city of what state/province of what country?

                  I think the OP's point is that if you are creating hardware for public consuption on pc's with the ability to run vista, $200 is not an unreasonable amount of money. It is also fair to say any price is unreasonable. I think MS feels like they need to spec out any driver in the interests of security, and they are the only trustworthy group to do that. In that vein, I am guessing they charge $200

                  • by tepples ( 727027 )

                    I think the OP's point is that if you are creating hardware for public consuption on pc's with the ability to run vista, $200 is not an unreasonable amount of money.

                    A file system is not hardware.

                    Nor is hardware developed for a close-knit circle of hobbyist hardware hackers exactly "for public consumption", but that's a topic for a different article.

                    • by tepples ( 727027 )

                      My main point was that if you are dropping a few hundred developer hours of time into writing and maintaining a file system driver, the $200 is insignificant.

                      Replace "a file system driver" with "an application" to see how Microsoft will try to kill free software. It has already happened to Xbox 360.

            • 64-bit versions of Windows Vista put an annoying "Test Mode" banner in all four corners of the desktop if the user installs a device driver that hasn't been signed by a publisher who pays an annual fee of at least $200 to a commercial certificate authority trusted by Microsoft.
              .

              $200 to certify a driver for something as elemental as a filing system seems reasonable enough.

              • by tepples ( 727027 )

                $200 to certify a driver for something as elemental as a filing system seems reasonable enough.

                The only reason that one would get away with it is because ext3fs (along with the rest of Linux) is GPLv2, not GPLv3. Under GPLv3, anybody who distributes copies of the installable file system would have to distribute Installation Information, and as I read the GPLv3, this would involve buying a certificate for each recipient of the source code.

            • I had to do the registry change to get an unsigned driver to load in 64-bit Vista Ultimate. The driver was for a Virtual CD/DVD drive so I could mount ISOs. I have never seen this banner you mention, and I have been running that machine for a year. In fact, I reloaded the machine once and had to install that same driver. Do you have a source for this, or did I get lucky? I would love to get a screenshot of it and set it as my co-worker's wallpaper.
      • Re: (Score:3, Informative)

        http://www.diskinternals.com/ [diskinternals.com] they have a freeware ext2/ext3 proggy called 'linux reader' ive had it installed for quite a few months. plays my ext3 mp3storage in winamp just fine.

    • Re:FAT32 (Score:5, Informative)

      by Ruie ( 30480 ) on Thursday August 14, 2008 @10:45PM (#24610111) Homepage

      I think he would have the same problem with a ntfs drive.

      The issue is that his Linux user setup and Windows user setup are different.
      So when he mounts the partition all files are owned by root (as shown on the screen), and some files have public permissions turned off - a reasonable thing.

      Thus what he needs to do is specify the owner of the files using uid=value
      option in /etc/fstab (uid value can be found via "getent passwd", it is numeric).

      For more info read "man mount" carefully.

      • Nope, FAT lacks the "owner" principle, the issue here is the 'System' bit. These are the file attributes FAT knows: - Read - Write - System
        • Re: (Score:2, Insightful)

          by RupW ( 515653 ) *

          These are the file attributes FAT knows:
          - Read
          - Write
          - System

          No, it's

          • r - read only
          • a - archive (set when the file is modified, i.e. can use as a simple 'needs backup' flag)
          • s - system
          • h - hidden
        • by Ruie ( 30480 )

          Nope, FAT lacks the "owner" principle, the issue here is the 'System' bit.

          These are the file attributes FAT knows:
          - Read
          - Write
          - System

          Yes, but when you mount FAT in Linux the files are assigned an owner - which is is either specified via options or, in their absence, is the user running mount.

    • NTFS-3G works pretty well. I'm not sure FAT32 is really necessary any more.

      Unless you have an SDHC card that you're sneakernetting between your PC and a digital camera, or you have an external hard drive that you're sneakernetting between a Windows or Ubuntu PC and either a Mac or an Xbox 360. Cameras, Macs, and game consoles tend not to work with NTFS out of the box.

    • Re: (Score:3, Informative)

      by macshit ( 157376 )

      NTFS-3G works pretty well. I'm not sure FAT32 is really necessary any more.

      FAT may suck, but it's the only thing understood by a lot of embedded software like BIOSes, device firmware, etc...

      Indeed, for that reason it seems like FAT may very well be more useful than NTFS. FAT will probably stay around for quite a while as a "braindead, but simple and widespread" exchange format, but the only excuse for NTFS is windows.

      • by Bert64 ( 520050 )

        Except that FAT craps out with large files, and things like dvd images are becoming increasingly common these days.
        Having a new universally supported FS would be good for everyone, but Microsoft will never support a third party fs unless absolutely forced to, and anything they make themselves will be closed and proprietary and thus useless as a universal transfer system.

        On the other hand, UFS is good, Linux, OSX, Solaris and BSD all support it out of the box (tho admittedly linux's support is quite poor).

      • by Ant P. ( 974313 )

        Why bother keeping FAT around? UDF is better in every way, and it's even more widespread.

      • On a whim, I checked for FAT64 and found it does exist (sort of) on Vista already. http://en.wikipedia.org/wiki/ExFAT [wikipedia.org]
    • Re: (Score:3, Informative)

      by Mattsson ( 105422 )

      I've had NTFS-G3 totally destroy two NTFS partitions with the Vista version of NTFS 3.1
      This seem to differ a bit from the XP version of NTFS 3.1

    • My experience with it says otherwise. I've had various issues with NTFS-3G when reading the hard disk on Windows. They are as follows:

      1) For some reason, when a linux app tries to save a file with a question mark ("?"), which is an INVALID character on Windows, NTFS-3G will allow it to continue. You can perfectly read the file on linux (and write back to it), but Windows complains.

      2) When using characters that have accents on them (like you see in romance languages - portuguese being my case), copying files

    • by Monsuco ( 998964 )

      NTFS-3G works pretty well. I'm not sure FAT32 is really necessary any more.

      NTFS-3G is good for Linux reading Windows partitions, but you can have it the other way too. Windows can be set to read and write [sourceforge.net] to Ext2/3 file systems. I imagine ext4 will work with this too when it comes out.

      • That depends. A normal ext4 partition can be mounted as an ext2 partition and read in Windows, like ext3 and ext2 partitions with the right software. However, if you mount an ext4 partition with extents support, then it can no longer be mounted as an ext2 partition, and then can't be read in Windows.

  • Damn (Score:5, Funny)

    by MichaelSmith ( 789609 ) on Thursday August 14, 2008 @09:58PM (#24609701) Homepage Journal
    For a moment there I thought somebody had fixed Ubuntu bug one [launchpad.net].
  • I suspsect that... (Score:4, Insightful)

    by Psychotria ( 953670 ) on Thursday August 14, 2008 @10:14PM (#24609853)
    if the owner/group permissions were set properly in fstab an easier solution would prevail
  • This isn't a "Vista" bug, as I've seen it happen frequently on a dual boot machine that is only XP+Ubuntu (no Vista)

    I ran into this not that long ago and was really stuck scratching my head for awhile, as the fstab settings were definitely correct. However, after a little "chmod -R" magic on the entire FAT32 partition, it reset the recalcitrant permissions and everything worked fine.

    • Re:Not a vista bug (Score:5, Informative)

      by Qzukk ( 229616 ) on Thursday August 14, 2008 @10:57PM (#24610199) Journal

      It's not a bug, it's old knowledge getting flushed out of the general awareness of the public. FAT has a read-only bit and Linux knows about it, it's in there along with the system and hidden file bits:

      #define ATTR_RO 1 /* read-only */

      (linux/msdos_fs.h)

      • Glad to see I'm wasn't the only one scratching my head about the claim that FAT32 doesn't support the read-only attribute.

        Damn kids these days, don't remember having to use ATTRIB...

        • by phorm ( 591458 )

          I seem to remember that being supported back as far as the old DOS days and FAT16.

          However, if it were a read-only set with something like "attrib", then one shouldn't be able to edit these files in windows either. Seems to be more an issue with the way the 'vfat' module interprets some FAT32 dirs.

        • Damn kids these days, don't remember having to use ATTRIB...

          I don't know what was better: Being 14 and using ATTRIB +H to "hide" my porn from my dad... or using ATTRIB -H to find his. ^_^

      • Look on the bright side. He might have posted that his large files got corrupted.
      • Kids these days with their fancy filesystems... The general computing knowledge of the current generation is woofully limited, oh well, better for us older fellas, more phish to catch. Hell, kids these days don't even know Vista still has a cli.... This is *NOT* a bug in Ubuntu or linux in general, it's Microsoft violating it's own filesystem spec to have neat tricks in a crappy OS
    • Re:Not a vista bug (Score:5, Informative)

      by greg1104 ( 461138 ) <gsmith@gregsmith.com> on Thursday August 14, 2008 @11:49PM (#24610565) Homepage

      This started in XP actually. The problem is that Microsoft sets the read-only attribute on the special folders that get custom views. See http://support.microsoft.com/kb/326549 [microsoft.com] for information about the root cause of the problem reported on this blog. Fixing it on the Windows side requires one to go all old-school and use attrib; cracked me up.

  • by Anonymous Coward on Thursday August 14, 2008 @10:20PM (#24609893)

    ... gets page linked from slashdot.

    Well, at least I adblock.

    • Yeah seriously, why the fuck is this on Slashdot? I'm not the "stuff that matters" whiner type but either timothy never used a Linux distro and thinks this is newsworthy, or this is the slowest news day ever :).
      • Re: (Score:1, Offtopic)

        by saturnism ( 177882 )

        that's it, i'm outa here. of all the years i stood by slashdot, this post just did it for me.

        good bye news for non nerds
        good bye dupes
        good bye first posts
        good bye trolls
        good bye horrible commenting system and the stupid slider

        • Re: (Score:2, Funny)

          by Anonymous Coward

          You obviously never really did fit in here. I mean, a true slashdotter would have titled his post "Last post!"

      • either timothy never used a Linux distro and thinks this is newsworthy, or this is the slowest news day ever

        Timothy was last seen putting Ubuntu on an XO. He's been using Linux at least since I met him in 1999.

        It's August, every day is a slow news day :)

      • Ha, I was so thinking the same thing. The gist of this "story" is that they had a problem getting Vista and Ubuntu to work together (*mock gasp* I've never heard of such a thing!) and then proceeded to fix it. *yawn* To top it all off the linked article is a blog post from the submitter. Give me a break.

      • by ROMRIX ( 912502 )

        Yeah seriously, why the fuck is this on Slashdot? I'm not the "stuff that matters" whiner type but either timothy never used a Linux distro and thinks this is newsworthy, or this is the slowest news day ever :).

        If you think it's slow now, just wait. Tomorrow we're discussing "Ubuntu, 'Where's root!'"

  • Apparently, there is a common belief that Ubuntu and Linux are the same. Actually, there are many types of Linux operating systems like Gentoo, Slackware, Fedora, or SUSE. So Ubuntu => Linux, but !(Linux => Ubuntu). For example, I have run into this issue (not bug) using Slackware, Gentoo, and I think also OpenBSD. My solution is easier, though: stay out of 'My Documents'.

    • But Ubuntu is the most used distros and it can be easily said that it is a Ubuntu bug rather than testing default installs of Gentoo/Slackware/SuSE/Fedora/Arch/etc.
      • well then I think Linus should mark every kernel bug as a specific ubuntu bug instead of testing all the distros, because obviously no one cares about the rest of them
        • But which is easier to test for someone who isn't a kernel hacker, A) a default Ubuntu or B) the latest kernel. But what about compile-time flags, who's kernel tree, etc. By basing it on a standard Ubuntu install, the average person can still report kernel bugs without messing around with kernel hacking.
      • For crying out loud, *THIS IS NOT A BUG*! Bloody idiots, take an old MS Dos manual and look up ATTRIB, *READ THE FUCKING MANUAL*
  • not news (Score:2, Informative)

    user discovers chmod ... blogs about it ... boooooring.
  • But this probably isn't that complicated. For year and years in strange situations or another, when files were placed on a fat32 partition in windows in certain conditions, users in Linux couldn't freely access the file unless the partition was mounted with "-o umask=000". In fact I just bought a shitty thumb drive, stuck it in my laptop running Slack and the hotplugging daemon (yes... slackware has that oooOOOOOoh) picked it up and mounted it. I opened it up and saw they had this E3 something or another wi

  • oh my god. (Score:5, Funny)

    by nawcom ( 941663 ) on Thursday August 14, 2008 @10:57PM (#24610195) Homepage

    The Effectiveness of the Ubuntu Forums [ubuntuforums.org]

    (The link this person gives in his blog post)

    I swear to christ, reading that page made me want to kill a kitten.

  • but there's no 'Mod article down'
  • What the fuck is the issue here? This is "normal" behaviour, Windows Vista sets the System bit (don't ask) on the directories, and that way they get mounted -ro for everyone but 'root'. Ever heard of the command "attrib" on DOS? Bug my ass.

"What man has done, man can aspire to do." -- Jerry Pournelle, about space flight

Working...