Debian bug report logs - #379 Repeatable mount(1) problem with removable disks Package: mount ; Reported by: Bill Mitchell ; 269 days old . ----------------------------------------------------------------------- Message received at debian-bugs: From pixar.com!bruce Fri Aug 4 21:30:44 1995 Return-Path: Received: from pixar.com by mongo.pixar.com with smtp (Smail3.1.28.1 #15) id m0seasm-000BCVC; Fri, 4 Aug 95 21:30 PDT Received: from mongo.pixar.com by pixar.com with SMTP id AA11083 (5.67b/IDA-1.5 for debian-bugs-pipe@mongo.pixar.com); Fri, 4 Aug 1995 21:30:39 -0700 Received: by mongo.pixar.com (Smail3.1.28.1 #15) id m0seasj-000BCVC; Fri, 4 Aug 95 21:30 PDT Message-Id: Date: Fri, 4 Aug 95 21:30 PDT From: bruce@pixar.com (Bruce Perens) To: mitchell@mdd.comm.mot.com, debian-bugs@pixar.com Subject: Re: Bug#379: Repeatable mount(1) problem with removable disks Bill, Can you report on whether or not this bug still exists in kernel-1.2.12 ? I consider this a kernel problem. Thanks Bruce Perens -- -- Attention Ham Radio Operators: For information on "Linux for Hams", read -- the World Wide Web page http://www.hams.com/perens/LinuxForHams, or send -- an e-mail message containing the word "help" to info@hams.com . ----------------------------------------------------------------------- Acknowledgement sent to bruce@pixar.com (Bruce Perens) : Extra info received and forwarded. Full text available. ----------------------------------------------------------------------- Information forwarded to debian-devel@pixar.com : Bug#379 ; Package mount . Full text available. ----------------------------------------------------------------------- Message received at debian-bugs: From mdd.comm.mot.com!mitchell Sun Feb 5 10:35:28 1995 Return-Path: Received: from pixar.com by mongo.pixar.com with smtp (Smail3.1.28.1 #15) id m0rbBny-0002u1C; Sun, 5 Feb 95 10:35 PST Received: from motgate.mot.com by pixar.com with SMTP id AA17392 (5.65c/IDA-1.4.4 for ); Sun, 5 Feb 1995 10:35:58 -0800 Received: from pobox.mot.com by motgate.mot.com with SMTP (5.67b/IDA-1.4.4/MOT-3.1 for ) id AA06029; Sun, 5 Feb 1995 12:35:22 -0600 Received: from mdd.comm.mot.com (mdisea.mdd.comm.mot.com) by pobox.mot.com with SMTP (5.67b/IDA-1.4.4/MOT-3.1) id AA21930; Sun, 5 Feb 1995 12:35:21 -0600 Received: from bb29c.mdd.comm.mot.com by mdd.comm.mot.com (4.1/SMI-4.1) id AA08919; Sun, 5 Feb 95 10:35:19 PST Received: by bb29c.mdd.comm.mot.com (4.1/SMI-4.1) id AA04239; Sun, 5 Feb 95 10:35:13 PST Date: Sun, 5 Feb 1995 10:35:12 -0800 (PST) From: Bill Mitchell X-Sender: mitchell@bb29c To: Ian Jackson Cc: debian-bugs@pixar.com Subject: Re: Bug#379: Repeatable mount(1) problem with removable disks In-Reply-To: Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII On Sun, 5 Feb 1995, Ian Jackson wrote: > [...] > There is an ioctl that causes the kernel to re-read the partition > table; this is used by fdisk. [... but not used by mount ... (wtm)] > > [...] what > if you insert a cartridge that has a /dev/sdc4 but whose sdc4 is the > wrong size ? Ouch! This sounds like I could have one cartridge with a single 150MB partition on /dev/sdc1 mounted, unmount it and mount /dev/scd1 from another cartridge which has two 75MB partitions, and trash the second partition. No... perhaps not. Surely mounting the new partition would involve reading in the superblock, or whatever disk usage and sizing information the mounted filesystem contains. On the other hand, what about the case where I had that 75MB /dev/sdc1 partition mounted, unmounted it and mounted a 150MB /dev/sdc1 partition from another cartridge? If the partition table is not reread, mount will read in disk usage information showing that the filesystem it's mounted is bigger than the un-reread partition table shows the partition where it resides to be. Messy. > The right solution is surely to make sure that the kernel spots the > cartridge change and invalidates the partition table information > itself. Yes, but is it "right" for mount not to be paranoid enough to re-read the partition table before attempting a mount? If nothing else, having mount reread the partition table before attempting a mount would provide protection to folks using a later vintage mount program with an earlier vintage kernel. > Is there anything about this in the SCSI HOWTO ? If not you should > probably report it as a bug to the SCSI mailing list. I happen have a removable-cartridge disk which has a SCSI interface, but it seems to me these same concerns would apply to a removeable cartridge disk with an IDE interface as well. ----------------------------------------------------------------------- Message sent: From: iwj@cam-orl.co.uk (Ian Jackson) To: Bill Mitchell Subject: Bug#379: Info received (was Bug#379: Repeatable mount(1) problem with removable disks) In-Reply-To: References: Thank you for the additional information you have supplied regarding your problem report. It has been forwarded to the developers to accompany the original report. If you wish to continue submit further information on your problem, please do the same thing again: send it to debian-bugs@pixar.com, ensuring that the Subject line starts with "Bug#379" or "Re: Bug#379" so that we can identify it as relating to the same problem. Please do not to reply to the address at the top of this message, unless you wish to report a problem with the bug-tracking system. Ian Jackson (maintainer, debian-bugs) ----------------------------------------------------------------------- Message sent to debian-devel@pixar.com: Subject: Bug#379: Repeatable mount(1) problem with removable disks Reply-To: Bill Mitchell , debian-bugs-done@pixar.com Resent-To: debian-devel@pixar.com Resent-From: Bill Mitchell Resent-Sender: iwj10@cus.cam.ac.uk Resent-Date: Sun, 05 Feb 1995 18:48:02 GMT Resent-Message-ID: X-Debian-PR-Package: mount X-Debian-PR-Keywords: Received: via spool for debian-bugs; Sun, 05 Feb 1995 18:48:02 GMT Received: with rfc822 via encapsulated-mail id 0205183636844; Sun, 05 Feb 1995 18:36:36 GMT Received: from pixar.com by mongo.pixar.com with smtp (Smail3.1.28.1 #15) id m0rbBny-0002u1C; Sun, 5 Feb 95 10:35 PST Received: from motgate.mot.com by pixar.com with SMTP id AA17392 (5.65c/IDA-1.4.4 for ); Sun, 5 Feb 1995 10:35:58 -0800 Received: from pobox.mot.com by motgate.mot.com with SMTP (5.67b/IDA-1.4.4/MOT-3.1 for ) id AA06029; Sun, 5 Feb 1995 12:35:22 -0600 Received: from mdd.comm.mot.com (mdisea.mdd.comm.mot.com) by pobox.mot.com with SMTP (5.67b/IDA-1.4.4/MOT-3.1) id AA21930; Sun, 5 Feb 1995 12:35:21 -0600 Received: from bb29c.mdd.comm.mot.com by mdd.comm.mot.com (4.1/SMI-4.1) id AA08919; Sun, 5 Feb 95 10:35:19 PST Received: by bb29c.mdd.comm.mot.com (4.1/SMI-4.1) id AA04239; Sun, 5 Feb 95 10:35:13 PST Date: Sun, 5 Feb 1995 10:35:12 -0800 (PST) From: Bill Mitchell X-Sender: mitchell@bb29c To: Ian Jackson Cc: debian-bugs@pixar.com In-Reply-To: Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII On Sun, 5 Feb 1995, Ian Jackson wrote: > [...] > There is an ioctl that causes the kernel to re-read the partition > table; this is used by fdisk. [... but not used by mount ... (wtm)] > > [...] what > if you insert a cartridge that has a /dev/sdc4 but whose sdc4 is the > wrong size ? Ouch! This sounds like I could have one cartridge with a single 150MB partition on /dev/sdc1 mounted, unmount it and mount /dev/scd1 from another cartridge which has two 75MB partitions, and trash the second partition. No... perhaps not. Surely mounting the new partition would involve reading in the superblock, or whatever disk usage and sizing information the mounted filesystem contains. On the other hand, what about the case where I had that 75MB /dev/sdc1 partition mounted, unmounted it and mounted a 150MB /dev/sdc1 partition from another cartridge? If the partition table is not reread, mount will read in disk usage information showing that the filesystem it's mounted is bigger than the un-reread partition table shows the partition where it resides to be. Messy. > The right solution is surely to make sure that the kernel spots the > cartridge change and invalidates the partition table information > itself. Yes, but is it "right" for mount not to be paranoid enough to re-read the partition table before attempting a mount? If nothing else, having mount reread the partition table before attempting a mount would provide protection to folks using a later vintage mount program with an earlier vintage kernel. > Is there anything about this in the SCSI HOWTO ? If not you should > probably report it as a bug to the SCSI mailing list. I happen have a removable-cartridge disk which has a SCSI interface, but it seems to me these same concerns would apply to a removeable cartridge disk with an IDE interface as well. ----------------------------------------------------------------------- Message received at debian-bugs: From cus.cam.ac.uk!iwj10 Sun Feb 5 08:52:42 1995 Return-Path: Received: from pixar.com by mongo.pixar.com with smtp (Smail3.1.28.1 #15) id m0rbACX-0006fTC; Sun, 5 Feb 95 08:52 PST Received: from bootes.cus.cam.ac.uk by pixar.com with SMTP id AA06611 (5.65c/IDA-1.4.4 for ); Sun, 5 Feb 1995 08:52:59 -0800 Received: by bootes.cus.cam.ac.uk (Smail-3.1.29.0 #24) id m0rbA9S-000C0QC; Sun, 5 Feb 95 16:49 GMT Received: by chiark (Smail3.1.28.1 #3) id m0rb6Fh-0002g8Z; Sun, 5 Feb 95 12:39 GMT Message-Id: Date: Sun, 5 Feb 95 12:39 GMT From: iwj10@cus.cam.ac.uk (Ian Jackson) To: debian-bugs@pixar.com Subject: Re: Bug#379: Repeatable mount(1) problem with removable disks Bill Mitchell writes ("Bug#379: Repeatable mount(1) problem with removable disks"): > [...] > I report this as a mount(1) problem instead of a kernel problem > because I've noticed that I can clear the problem by running > fdisk /dev/sdc > and then exiting fdisk. If I do this in between the unmount of a > cartridge with one partition present and the mount of a cartridge > with another partition present, the mount succeeds. Apparently > fdisk(1) causes the partition information to be re-read from the > disk. There is an ioctl that causes the kernel to re-read the partition table; this is used by fdisk. > I'm suggesting that mount(1) should also re-read the partition > information from the disk. It should either do this all the time, > or it should do it after a mount failure, then retrythe mount request > before considering the failure to be a hard one and reporting it. No, I don't think it should. Certainly the latter won't work - what if you insert a cartridge that has a /dev/sdc4 but whose sdc4 is the wrong size ? The right solution is surely to make sure that the kernel spots the cartridge change and invalidates the partition table information itself. Is there anything about this in the SCSI HOWTO ? If not you should probably report it as a bug to the SCSI mailing list. Regards, Ian. ----------------------------------------------------------------------- Message sent: From: iwj@cam-orl.co.uk (Ian Jackson) To: iwj10@cus.cam.ac.uk (Ian Jackson) Subject: Bug#379: Info received (was Bug#379: Repeatable mount(1) problem with removable disks) In-Reply-To: References: Thank you for the additional information you have supplied regarding your problem report. It has been forwarded to the developers to accompany the original report. If you wish to continue submit further information on your problem, please do the same thing again: send it to debian-bugs@pixar.com, ensuring that the Subject line starts with "Bug#379" or "Re: Bug#379" so that we can identify it as relating to the same problem. Please do not to reply to the address at the top of this message, unless you wish to report a problem with the bug-tracking system. Ian Jackson (maintainer, debian-bugs) ----------------------------------------------------------------------- Message sent to debian-devel@pixar.com: Subject: Bug#379: Repeatable mount(1) problem with removable disks Reply-To: iwj10@cus.cam.ac.uk (Ian Jackson), debian-bugs-done@pixar.com Resent-To: debian-devel@pixar.com Resent-From: iwj10@cus.cam.ac.uk (Ian Jackson) Resent-Sender: iwj10@cus.cam.ac.uk Resent-Date: Sun, 05 Feb 1995 17:03:04 GMT Resent-Message-ID: X-Debian-PR-Package: mount X-Debian-PR-Keywords: Received: via spool for debian-bugs; Sun, 05 Feb 1995 17:03:04 GMT Received: with rfc822 via encapsulated-mail id 020516571324976; Sun, 05 Feb 1995 16:57:13 GMT Received: from pixar.com by mongo.pixar.com with smtp (Smail3.1.28.1 #15) id m0rbACX-0006fTC; Sun, 5 Feb 95 08:52 PST Received: from bootes.cus.cam.ac.uk by pixar.com with SMTP id AA06611 (5.65c/IDA-1.4.4 for ); Sun, 5 Feb 1995 08:52:59 -0800 Received: by bootes.cus.cam.ac.uk (Smail-3.1.29.0 #24) id m0rbA9S-000C0QC; Sun, 5 Feb 95 16:49 GMT Received: by chiark (Smail3.1.28.1 #3) id m0rb6Fh-0002g8Z; Sun, 5 Feb 95 12:39 GMT Message-Id: Date: Sun, 5 Feb 95 12:39 GMT From: iwj10@cus.cam.ac.uk (Ian Jackson) To: debian-bugs@pixar.com Bill Mitchell writes ("Bug#379: Repeatable mount(1) problem with removable disks"): > [...] > I report this as a mount(1) problem instead of a kernel problem > because I've noticed that I can clear the problem by running > fdisk /dev/sdc > and then exiting fdisk. If I do this in between the unmount of a > cartridge with one partition present and the mount of a cartridge > with another partition present, the mount succeeds. Apparently > fdisk(1) causes the partition information to be re-read from the > disk. There is an ioctl that causes the kernel to re-read the partition table; this is used by fdisk. > I'm suggesting that mount(1) should also re-read the partition > information from the disk. It should either do this all the time, > or it should do it after a mount failure, then retrythe mount request > before considering the failure to be a hard one and reporting it. No, I don't think it should. Certainly the latter won't work - what if you insert a cartridge that has a /dev/sdc4 but whose sdc4 is the wrong size ? The right solution is surely to make sure that the kernel spots the cartridge change and invalidates the partition table information itself. Is there anything about this in the SCSI HOWTO ? If not you should probably report it as a bug to the SCSI mailing list. Regards, Ian. ----------------------------------------------------------------------- Message received at debian-bugs: From mdd.comm.mot.com!mitchell Sat Feb 4 10:45:20 1995 Return-Path: Received: from pixar.com by mongo.pixar.com with smtp (Smail3.1.28.1 #15) id m0rapTz-0002piC; Sat, 4 Feb 95 10:45 PST Received: from motgate.mot.com by pixar.com with SMTP id AA24564 (5.65c/IDA-1.4.4 for ); Sat, 4 Feb 1995 10:45:54 -0800 Received: from pobox.mot.com by motgate.mot.com with SMTP (5.67b/IDA-1.4.4/MOT-3.1 for ) id AA28632; Sat, 4 Feb 1995 12:45:18 -0600 Received: from mdd.comm.mot.com (mdisea.mdd.comm.mot.com) by pobox.mot.com with SMTP (5.67b/IDA-1.4.4/MOT-3.1 for ) id AA05388; Sat, 4 Feb 1995 12:45:17 -0600 Received: from bb29c.mdd.comm.mot.com by mdd.comm.mot.com (4.1/SMI-4.1) id AA01421; Sat, 4 Feb 95 10:45:16 PST Received: by bb29c.mdd.comm.mot.com (4.1/SMI-4.1) id AA03215; Sat, 4 Feb 95 10:45:14 PST Date: Sat, 4 Feb 1995 10:45:13 -0800 (PST) From: Bill Mitchell X-Sender: mitchell@bb29c To: debian-bugs@pixar.com Subject: Repeatable mount(1) problem with removable disks In-Reply-To: Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII PACKAGE: mount VARSION: 1.10 PACKAGE_REVISION: 4 (and later?) One of the devices on my system is a removeable cartridge disk drive. I have this on /dev/sdc. The disk cartridges come preformatted for MSDOS, with partitions 1 thru 3 not present and with an MSDOS filesystem in partition 4. I use them the way they come out of the box for MSDOS use. I also use this disk for ext2 filesystems. I usually repartition a disk cartridge so that partition 1 is present instead of partition 4, and put an ext2 filesystem in partition 1. So, to mount a disk cartridge, I usually say one of the following: mount -t msdos /dev/sdc4 /mnt mount -t ext2 /dev/sdc1 /mnt This works fine for the initial mount of this disk, and works just fine if I unmount and remount disks of the same type. However, if I unmount an MSDOS disk and try to mount an ext2 disk, mount(1) fails with the message: mount: /dev/sdc1 is not a valid block device and I get a similar failure message for /dev/sdc4 when trying to mount an MSDOS disk after I've unmounted an ext2 disk. Apparently the partition table info is kept cached somewhere in the kernel, and mount requests use this cached info instead of re-reading the partition table from the disk. I report this as a mount(1) problem instead of a kernel problem because I've noticed that I can clear the problem by running fdisk /dev/sdc and then exiting fdisk. If I do this in between the unmount of a cartridge with one partition present and the mount of a cartridge with another partition present, the mount succeeds. Apparently fdisk(1) causes the partition information to be re-read from the disk. I'm suggesting that mount(1) should also re-read the partition information from the disk. It should either do this all the time, or it should do it after a mount failure, then retrythe mount request before considering the failure to be a hard one and reporting it. mitchell@mdd.comm.mot.com (Bill Mitchell) ----------------------------------------------------------------------- Message sent: From: iwj@cam-orl.co.uk (Ian Jackson) To: Bill Mitchell Subject: Bug#379: Acknowledgement (was: Repeatable mount(1) problem with removable disks) In-Reply-To: References: Thank you for the problem report you have sent regarding Debian GNU/Linux. This is an automatically generated reply, to let you know your message has been received. It is being forwarded to the developers' mailing list for their attention; they will reply in due course. If you wish to submit further information on your problem, please send it to debian-bugs@pixar.com, but please ensure that the Subject line of your message starts with "Bug#379" or "Re: Bug#379" so that we can identify it as relating to the same problem. Please do not to reply to the address at the top of this message, unless you wish to report a problem with the bug-tracking system. Ian Jackson (maintainer, debian-bugs) ----------------------------------------------------------------------- Message sent to debian-devel@pixar.com: Subject: Bug#379: Repeatable mount(1) problem with removable disks Reply-To: Bill Mitchell , debian-bugs-done@pixar.com Resent-To: debian-devel@pixar.com Resent-From: Bill Mitchell Resent-Sender: iwj10@cus.cam.ac.uk Resent-Date: Sat, 04 Feb 1995 18:48:05 GMT Resent-Message-ID: X-Debian-PR-Package: mount X-Debian-PR-Keywords: Received: via spool for debian-bugs; Sat, 04 Feb 1995 18:48:05 GMT Received: with rfc822 via encapsulated-mail id 02041846278127; Sat, 04 Feb 1995 18:46:27 GMT Received: from pixar.com by mongo.pixar.com with smtp (Smail3.1.28.1 #15) id m0rapTz-0002piC; Sat, 4 Feb 95 10:45 PST Received: from motgate.mot.com by pixar.com with SMTP id AA24564 (5.65c/IDA-1.4.4 for ); Sat, 4 Feb 1995 10:45:54 -0800 Received: from pobox.mot.com by motgate.mot.com with SMTP (5.67b/IDA-1.4.4/MOT-3.1 for ) id AA28632; Sat, 4 Feb 1995 12:45:18 -0600 Received: from mdd.comm.mot.com (mdisea.mdd.comm.mot.com) by pobox.mot.com with SMTP (5.67b/IDA-1.4.4/MOT-3.1 for ) id AA05388; Sat, 4 Feb 1995 12:45:17 -0600 Received: from bb29c.mdd.comm.mot.com by mdd.comm.mot.com (4.1/SMI-4.1) id AA01421; Sat, 4 Feb 95 10:45:16 PST Received: by bb29c.mdd.comm.mot.com (4.1/SMI-4.1) id AA03215; Sat, 4 Feb 95 10:45:14 PST Date: Sat, 4 Feb 1995 10:45:13 -0800 (PST) From: Bill Mitchell X-Sender: mitchell@bb29c To: debian-bugs@pixar.com In-Reply-To: Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII PACKAGE: mount VARSION: 1.10 PACKAGE_REVISION: 4 (and later?) One of the devices on my system is a removeable cartridge disk drive. I have this on /dev/sdc. The disk cartridges come preformatted for MSDOS, with partitions 1 thru 3 not present and with an MSDOS filesystem in partition 4. I use them the way they come out of the box for MSDOS use. I also use this disk for ext2 filesystems. I usually repartition a disk cartridge so that partition 1 is present instead of partition 4, and put an ext2 filesystem in partition 1. So, to mount a disk cartridge, I usually say one of the following: mount -t msdos /dev/sdc4 /mnt mount -t ext2 /dev/sdc1 /mnt This works fine for the initial mount of this disk, and works just fine if I unmount and remount disks of the same type. However, if I unmount an MSDOS disk and try to mount an ext2 disk, mount(1) fails with the message: mount: /dev/sdc1 is not a valid block device and I get a similar failure message for /dev/sdc4 when trying to mount an MSDOS disk after I've unmounted an ext2 disk. Apparently the partition table info is kept cached somewhere in the kernel, and mount requests use this cached info instead of re-reading the partition table from the disk. I report this as a mount(1) problem instead of a kernel problem because I've noticed that I can clear the problem by running fdisk /dev/sdc and then exiting fdisk. If I do this in between the unmount of a cartridge with one partition present and the mount of a cartridge with another partition present, the mount succeeds. Apparently fdisk(1) causes the partition information to be re-read from the disk. I'm suggesting that mount(1) should also re-read the partition information from the disk. It should either do this all the time, or it should do it after a mount failure, then retrythe mount request before considering the failure to be a hard one and reporting it. mitchell@mdd.comm.mot.com (Bill Mitchell) ----------------------------------------------------------------------- Ian Jackson / iwj10@thor.cam.ac.uk , with the debian-bugs tracking mechanism This page last modified 07:43:01 GMT Wed 01 Nov