Re: [Cooker] [Bug 2846] [mtools] Floppy formatter incompatible with supermount
On Tuesday 29 July 2003 11:00, François Pons wrote: > > > > tell Juan to include my supermount to make it possible :) > > Ah, does it means the current kernel doesn't support this functionnalities > ? Well, I have been trying to harrass Juan about this for a few months now. It would probably fix all the supermount bugs in bugzilla in one sweep. Plus it gives us some nice features like this one. But for some of these features apps have to be adepted. So we really need it a considerable time before a freeze. Juan, please? Patch for latest kernel is already done for you... d.
Re: [Cooker] [Bug 2846] [mtools] Floppy formatter incompatible with supermount
On Tue, 29 Jul 2003, Eric Fernandez wrote: > Hi > > and what about using supermount-ng ? ;) > > Zeb we were talking about that: supermount-ng==andrey's supermount d. > >
Re: [Cooker] [Bug 2846] [mtools] Floppy formatter incompatible with supermount
On Wednesday 30 July 2003 14:07, François Pons wrote: > "Andrey Borzenkov" <[EMAIL PROTECTED]> writes: > > > Can you make /proc/fs/supermount/subfs owner the user who has control > > > over the floppy device ? > > > > the simplest way is to use pam_console to reassign ownership. > > I'll get a look. will it be OK? > it needs even more patches in main kernel so I'd rather avoid it. What is wrong with this trivial helper (attached) and usermode: pts/2}% subfs --help Unknown option: help Usage: /usr/sbin/subfs {options ...} ... -e, --enable: enable subfs -d, --disable: disable subfs -r, --release: umount subfs -f, --force:force umount if subfs busy -s, --status: show subfs status {pts/2}% LC_ALL=C ll =subfs lrwxrwxrwx1 root root 13 Mar 23 13:52 /usr//bin/subfs -> consolehelper* {pts/2}% LC_ALL=C ll /etc/security/console.apps/subfs -rw-r--r--1 root root0 Mar 30 19:12 /etc/security/console.apps/subfs {pts/2}% cat /etc/pam.d/subfs #%PAM-1.0 auth sufficient /lib/security/pam_rootok.so auth required /lib/security/pam_console.so #auth required/lib/security/pam_stack.so service=system-auth accountrequired /lib/security/pam_permit.so I guess it is more flexible than hardcoding it in kernel. subfs Description: Perl program
Re: [Cooker] [Bug 2846] [mtools] Floppy formatter incompatible with supermount
"Andrey Borzenkov" <[EMAIL PROTECTED]> writes: > > Can you make /proc/fs/supermount/subfs owner the user who has control over the > > floppy device ? > > the simplest way is to use pam_console to reassign ownership. > I'll get a look. will it be OK? Yes, but I didn't checked if it was not working, but since it is a big problem if trying to umount the device, it will be much nicer to have this functionnalities. And yes, it still need to parse /etc/mtab (or maybe if the format fail, trying to unlock the device and re-opens ?). > > It is maybe already the case but it will be much safer to allow > > the same user and only him to release the supermounted floppy. > > > > "the same" - which one? I could add mout option to specify uid/gid > but is it not bloat? > > I suggest trying pam_console first, if we hit problems. let's solve > them. /proc/fs/supermount/subfs is available as soon as supermount > modules is loaded so it does not depend on mounting/unmounting. Yes, in fact I didn't check but checked umount was not working in such case. François.
Re: [Cooker] [Bug 2846] [mtools] Floppy formatter incompatible with supermount
-Original Message- > > "Andrey Borzenkov" <[EMAIL PROTECTED]> writes: > > > > --- Additional Comments From [EMAIL PROTECTED] 2003-28-07 19:11 --- > > > Yes, a setuid wrapper could be made for exemple to solve the problem. > > > > > I don't know if supermount can be modified safely to release floppy as soon as > > > an application try to open /dev/fd0 in writing, but it should be possible too > > > and it could be the best place to fix the problem. > > > > supermount can release device (and block further access to it); > > you have to do it manually using /proc interface. > > > > doing this automatically involves too many cnages in kernel and I > > won't do it; adding wrapper that does > > > > echo /dev/fd0 release disable > /proc/fs/supermount/subfs > > > > to GNOME formatter is far more simple. > > Yes, I agree, I didn't want to parse /etc/mtab and do an umount (in root, or > practically setuid root script) if the device was mounted. > you still have to parse something to know it is supermounted and possibly busy. of course it is possible to "just" attempt to disable it, still it probably have to warn user if subfs is currently mounted. > > > Note: formating a floppy while mounted is impossible, the device is busy, there > > > is nothing to do here unless umounting the floppy (in root furthermore when > > > using supermount). > > > > do not use real device name when mounting supermount; use "none". Then > > nobody can ever check that /dev/fd0 is mounted by supermount. > > > > use /proc interface to release it before formatting. > > > > tell Juan to include my supermount to make it possible :) > > Ah, does it means the current kernel doesn't support this functionnalities ? > 2.6 kernel package in cooker does AFAIK :) > Can you make /proc/fs/supermount/subfs owner the user who has control over the > floppy device ? the simplest way is to use pam_console to reassign ownership. I'll get a look. will it be OK? else it seems possible; would something like capable(CAP_SYS_RAWIO) || sys_access("/dev/fd0", W_OK) do or we need better capability? > It is maybe already the case but it will be much safer to allow > the same user and only him to release the supermounted floppy. > "the same" - which one? I could add mout option to specify uid/gid but is it not bloat? I suggest trying pam_console first, if we hit problems. let's solve them. /proc/fs/supermount/subfs is available as soon as supermount modules is loaded so it does not depend on mounting/unmounting. > FranГois. >
Re: [Cooker] [Bug 2846] [mtools] Floppy formatter incompatible with supermount
Buchan Milne <[EMAIL PROTECTED]> writes: > FACORAT Fabrice wrote: > > > Maybe Juan is working on a better version ? > > There is no better version. > > > In fact I don't know ( and Juan silence concerning this is disturbing ), > > but what I know is that we need a reliable and fully functionnal > > supermount, > > That is what supermount-ng is. > > > or else we should better remove it and use autofs. > > Autofs has worse problems for removable media IMHO. > > Please, try a kernel with supermount-ng first, you will see it works > much better. I updated bug 2846 to reflects this need to supermount-ng = Andrej supermount modified version more clearly. François.
Re: [Cooker] [Bug 2846] [mtools] Floppy formatter incompatible with supermount
On Tue, 29 Jul 2003, Eric Fernandez wrote: > Hi > > and what about using supermount-ng ? ;) > > Zeb we were talking about that: supermount-ng==andrey's supermount d. > >
Re: [Cooker] [Bug 2846] [mtools] Floppy formatter incompatible with supermount
Hi and what about using supermount-ng ? ;) Zeb
Re: [Cooker] [Bug 2846] [mtools] Floppy formatter incompatible with supermount
On Tuesday 29 July 2003 11:00, François Pons wrote: > > > > tell Juan to include my supermount to make it possible :) > > Ah, does it means the current kernel doesn't support this functionnalities > ? Well, I have been trying to harrass Juan about this for a few months now. It would probably fix all the supermount bugs in bugzilla in one sweep. Plus it gives us some nice features like this one. But for some of these features apps have to be adepted. So we really need it a considerable time before a freeze. Juan, please? Patch for latest kernel is already done for you... d.
Re: [Cooker] [Bug 2846] [mtools] Floppy formatter incompatible with supermount
"Andrey Borzenkov" <[EMAIL PROTECTED]> writes: > > --- Additional Comments From [EMAIL PROTECTED] 2003-28-07 19:11 --- > > Yes, a setuid wrapper could be made for exemple to solve the problem. > > > I don't know if supermount can be modified safely to release floppy as soon as > > an application try to open /dev/fd0 in writing, but it should be possible too > > and it could be the best place to fix the problem. > > supermount can release device (and block further access to it); > you have to do it manually using /proc interface. > > doing this automatically involves too many cnages in kernel and I > won't do it; adding wrapper that does > > echo /dev/fd0 release disable > /proc/fs/supermount/subfs > > to GNOME formatter is far more simple. Yes, I agree, I didn't want to parse /etc/mtab and do an umount (in root, or practically setuid root script) if the device was mounted. > > Note: formating a floppy while mounted is impossible, the device is busy, there > > is nothing to do here unless umounting the floppy (in root furthermore when > > using supermount). > > do not use real device name when mounting supermount; use "none". Then > nobody can ever check that /dev/fd0 is mounted by supermount. > > use /proc interface to release it before formatting. > > tell Juan to include my supermount to make it possible :) Ah, does it means the current kernel doesn't support this functionnalities ? Can you make /proc/fs/supermount/subfs owner the user who has control over the floppy device ? It is maybe already the case but it will be much safer to allow the same user and only him to release the supermounted floppy. François.