Skip to contentSkip to conversation

btrfs-progs: require mkfs -f force option to overwrite filesystem or partition table

RFC PATCH In review Eric Sandeen 9 participants 17 messages Feb 14, 2013 – Feb 28, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 btrfs-progs: require mkfs -f force option to overwrite filesystem or partition table

The core of this is shamelessly stolen from xfsprogs.

Use blkid to detect an existing filesystem or partition table on any of the target devices. If something is found, require the '-f' option to overwrite it, hopefully avoiding disaster due to mistyped devicenames, etc.

# mkfs.btrfs /dev/sda1

WARNING! - Btrfs v0.20-rc1-59-gd00279c-dirty IS EXPERIMENTAL WARNING! - see http://btrfs.wiki.kernel.org before using

/dev/sda1 appears to contain an existing filesystem (xfs). Use the -f option to force overwrite.
#

This does introduce a requirement on libblkid.

---

Note: this depends on my earlier small series, [PATCH 1/2] btrfs-progs: fix mkfs.btrfs -r option [PATCH 2/2, RFC] btrfs-progs: overhaul mkfs.btrfs -r option

Makefile +1 -1
index 4894903..bd0e8a6 100644
@@ -17,7 +17,7 @@ DEPFLAGS = -Wp,-MMD,$(@D)/.$(@F).d,-MT,$@
 INSTALL = install
 prefix ?= /usr/local
 bindir = $(prefix)/bin
-LIBS=-luuid -lm
+LIBS=-luuid -lblkid -lm
 RESTORE_LIBS=-lz
 
 progs = btrfsctl mkfs.btrfs btrfs-debug-tree btrfs-show btrfs-vol btrfsck \
man/mkfs.btrfs.8.in +6 -0
index c9f9e4f..9188201 100644
@@ -6,6 +6,7 @@ mkfs.btrfs \- create a btrfs filesystem
 [ \fB\-A\fP\fI alloc-start\fP ]
 [ \fB\-b\fP\fI byte-count\fP ]
 [ \fB\-d\fP\fI data-profile\fP ]
+[ \fB\-f\fP\fI ]
 [ \fB\-l\fP\fI leafsize\fP ]
 [ \fB\-L\fP\fI label\fP ]
 [ \fB\-m\fP\fI metadata profile\fP ]
@@ -38,6 +39,11 @@ mkfs.btrfs uses all the available storage for the filesystem.
 Specify how the data must be spanned across the devices specified. Valid
 values are raid0, raid1, raid10 or single.
 .TP
+\fB\-f\fR
+Force overwrite when an existing filesystem is detected on the device.
+By default, mkfs.btrfs will not write to the device if it suspects that 
+there is a filesystem or partition table on the device already.
+.TP
 \fB\-l\fR, \fB\-\-leafsize \fIsize\fR
 Specify the leaf size, the least data item in which btrfs stores data. The
 default value is the page size.
mkfs.c +100 -1

143 lines not on this page · raw message

#1 lore raw
Quoted 6 lines · Eric Sandeen

The core of this is shamelessly stolen from xfsprogs.

Use blkid to detect an existing filesystem or partition table on any of the target devices. If something is found, require the '-f' option to overwrite it, hopefully avoiding disaster due to mistyped devicenames, etc.

So, this has always been the one thing I didn't want to steal from XFS. I don't have a good reason though, and it seems like it is time to bring this in.

Just to get back at you though, I'll turn on an incandescent light bulb every time I have to use -f.

-chris

#2 lore raw
Quoted 11 lines · Chris Mason
Quoted 6 lines · Eric Sandeen

The core of this is shamelessly stolen from xfsprogs.

Use blkid to detect an existing filesystem or partition table on any of the target devices. If something is found, require the '-f' option to overwrite it, hopefully avoiding disaster due to mistyped devicenames, etc.

So, this has always been the one thing I didn't want to steal from XFS. I don't have a good reason though, and it seems like it is time to bring this in.

If you really don't want to, I'm not wedded to it. It just came up on the other thread, and it's easy enough to throw it out there for discussion.

I don't like a million "are you really sure?!"'s from tools, but now and then it's nice to have a little barrier to screwup.

So, do as you like with it, I won't be offended if you don't want it.

Quoted 2 lines

Just to get back at you though, I'll turn on an incandescent light bulb every time I have to use -f.

NOOOO that seals the deal. I retract the patch. ;)

-Eric

Quoted 2 lines

-chris

#3 lore raw
In reply to Eric Sandeen
Quoted 12 lines · Eric Sandeen

The core of this is shamelessly stolen from xfsprogs.

Use blkid to detect an existing filesystem or partition table on any of the target devices. If something is found, require the '-f' option to overwrite it, hopefully avoiding disaster due to mistyped devicenames, etc.

# mkfs.btrfs /dev/sda1

WARNING! - Btrfs v0.20-rc1-59-gd00279c-dirty IS EXPERIMENTAL WARNING! - see http://btrfs.wiki.kernel.org before using

Quoted 7 lines

/dev/sda1 appears to contain an existing filesystem (xfs). Use the -f option to force overwrite.
#

This does introduce a requirement on libblkid.

This means that it is now required to change all occurrences of "mkfs.btrfs" to "mkfs.btrfs -f" everywhere. Can't we first establish a time period of 100 years where the -f option is tolerated and ignored, and then in 2113 we require that the users add the -f option?

(Just had to do this string replacement everywhere, and had to add -f to xfstest's _scratch_mkfs in common.rc as well). Sigh.

#4 lore raw
Quoted 17 lines · Stefan Behrens
Quoted 16 lines · Eric Sandeen

The core of this is shamelessly stolen from xfsprogs.

Use blkid to detect an existing filesystem or partition table on any of the target devices. If something is found, require the '-f' option to overwrite it, hopefully avoiding disaster due to mistyped devicenames, etc.

# mkfs.btrfs /dev/sda1

WARNING! - Btrfs v0.20-rc1-59-gd00279c-dirty IS EXPERIMENTAL WARNING! - see http://btrfs.wiki.kernel.org before using

/dev/sda1 appears to contain an existing filesystem (xfs). Use the -f option to force overwrite.
#

Quoted 6 lines
Quoted 3 lines

This does introduce a requirement on libblkid.

This means that it is now required to change all occurrences of "mkfs.btrfs" to "mkfs.btrfs -f" everywhere. Can't we first establish a

I also think so.
It means -f is not significant to me, I think.
(Most of my test scripts fails without -f. So I'll always type "mkfs.btrfs -f")

Therefore I want you to revert commit:2a2d8e1962e8b6cda7b0a7584f6d2fb95d442cb6.
btrfs-progs: require mkfs -f force option to overwrite filesystem or partition table

How do you think about it?

Thanks,
Tsutomu

Quoted 6 lines

time period of 100 years where the -f option is tolerated and ignored, and then in 2113 we require that the users add the -f option?

(Just had to do this string replacement everywhere, and had to add -f to xfstest's _scratch_mkfs in common.rc as well). Sigh.

#5 lore raw
Quoted 25 lines · Tsutomu Itoh
Quoted 17 lines · Stefan Behrens
Quoted 16 lines · Eric Sandeen

The core of this is shamelessly stolen from xfsprogs.

Use blkid to detect an existing filesystem or partition table on any of the target devices. If something is found, require the '-f' option to overwrite it, hopefully avoiding disaster due to mistyped devicenames, etc.

# mkfs.btrfs /dev/sda1

WARNING! - Btrfs v0.20-rc1-59-gd00279c-dirty IS EXPERIMENTAL WARNING! - see http://btrfs.wiki.kernel.org before using

/dev/sda1 appears to contain an existing filesystem (xfs). Use the -f option to force overwrite.
#

Quoted 6 lines
Quoted 3 lines

This does introduce a requirement on libblkid.

This means that it is now required to change all occurrences of "mkfs.btrfs" to "mkfs.btrfs -f" everywhere. Can't we first establish a

Quoted 8 lines

I also think so.
It means -f is not significant to me, I think.
(Most of my test scripts fails without -f. So I'll always type "mkfs.btrfs -f")

Therefore I want you to revert commit:2a2d8e1962e8b6cda7b0a7584f6d2fb95d442cb6.
btrfs-progs: require mkfs -f force option to overwrite filesystem or partition table

How do you think about it?

What if you submit a patch to look at an environment variable, BTRFS_CLOBBERS_ALL=1 which causes it to not require -f to overwrite? Then you can just set it once at the top of your test environment, and not change every instance?

Otherwise, I guess I think:

WARNING! - Btrfs v0.20-rc1-212-gf6ef8b5 IS EXPERIMENTAL

and we need to expect that things might change ...

-Eric

Quoted 10 lines

Thanks,
Tsutomu

Quoted 6 lines

time period of 100 years where the -f option is tolerated and ignored, and then in 2113 we require that the users add the -f option?

(Just had to do this string replacement everywhere, and had to add -f to xfstest's _scratch_mkfs in common.rc as well). Sigh.

#6 lore raw
Quoted 33 lines · Eric Sandeen
Quoted 32 lines · Tsutomu Itoh
Quoted 17 lines · Stefan Behrens
Quoted 16 lines · Eric Sandeen

The core of this is shamelessly stolen from xfsprogs.

Use blkid to detect an existing filesystem or partition table on any of the target devices. If something is found, require the '-f' option to overwrite it, hopefully avoiding disaster due to mistyped devicenames, etc.

# mkfs.btrfs /dev/sda1

WARNING! - Btrfs v0.20-rc1-59-gd00279c-dirty IS EXPERIMENTAL WARNING! - see http://btrfs.wiki.kernel.org before using

/dev/sda1 appears to contain an existing filesystem (xfs). Use the -f option to force overwrite.
#

Quoted 6 lines
Quoted 3 lines

This does introduce a requirement on libblkid.

This means that it is now required to change all occurrences of "mkfs.btrfs" to "mkfs.btrfs -f" everywhere. Can't we first establish a

I also think so.
It means -f is not significant to me, I think.
(Most of my test scripts fails without -f. So I'll always type "mkfs.btrfs -f")

Therefore I want you to revert commit:2a2d8e1962e8b6cda7b0a7584f6d2fb95d442cb6.
btrfs-progs: require mkfs -f force option to overwrite filesystem or partition table

Quoted 6 lines
Quoted 1 line

How do you think about it?

What if you submit a patch to look at an environment variable, BTRFS_CLOBBERS_ALL=1 which causes it to not require -f to overwrite? Then you can just set it once at the top of your test environment, and not change every instance?

Yes. But,

Quoted 1 line
Quoted 1 line

(Most of my test scripts fails without -f. So I'll always type "mkfs.btrfs -f")

is one example.

Almost everyone types "mkfs.btrfs -f" (or BTRFS_CLOBBERS_ALL=1 :) unconditionally, I think.
So, I think -f option is almost meaningless.

Quoted 5 lines

Otherwise, I guess I think:

WARNING! - Btrfs v0.20-rc1-212-gf6ef8b5 IS EXPERIMENTAL

and we need to expect that things might change ...

EXPERIMENTAL... It's certainly so.
However, I think that we should not add the option that it troubles a lot of people.

Thanks,
Tsutomu

Quoted 14 lines

-Eric

Quoted 10 lines

Thanks,
Tsutomu

Quoted 6 lines

time period of 100 years where the -f option is tolerated and ignored, and then in 2113 we require that the users add the -f option?

(Just had to do this string replacement everywhere, and had to add -f to xfstest's _scratch_mkfs in common.rc as well). Sigh.

#7 lore raw
Quoted 51 lines · Tsutomu Itoh
Quoted 33 lines · Eric Sandeen
Quoted 32 lines · Tsutomu Itoh
Quoted 17 lines · Stefan Behrens
Quoted 16 lines · Eric Sandeen

The core of this is shamelessly stolen from xfsprogs.

Use blkid to detect an existing filesystem or partition table on any of the target devices. If something is found, require the '-f' option to overwrite it, hopefully avoiding disaster due to mistyped devicenames, etc.

# mkfs.btrfs /dev/sda1

WARNING! - Btrfs v0.20-rc1-59-gd00279c-dirty IS EXPERIMENTAL WARNING! - see http://btrfs.wiki.kernel.org before using

/dev/sda1 appears to contain an existing filesystem (xfs). Use the -f option to force overwrite.
#

Quoted 6 lines
Quoted 3 lines

This does introduce a requirement on libblkid.

This means that it is now required to change all occurrences of "mkfs.btrfs" to "mkfs.btrfs -f" everywhere. Can't we first establish a

I also think so.
It means -f is not significant to me, I think.
(Most of my test scripts fails without -f. So I'll always type "mkfs.btrfs -f")

Therefore I want you to revert commit:2a2d8e1962e8b6cda7b0a7584f6d2fb95d442cb6.
btrfs-progs: require mkfs -f force option to overwrite filesystem or partition table

Quoted 6 lines
Quoted 1 line

How do you think about it?

What if you submit a patch to look at an environment variable, BTRFS_CLOBBERS_ALL=1 which causes it to not require -f to overwrite? Then you can just set it once at the top of your test environment, and not change every instance?

Yes. But,

Quoted 1 line
Quoted 1 line

(Most of my test scripts fails without -f. So I'll always type "mkfs.btrfs -f")

is one example.

Almost everyone types "mkfs.btrfs -f" (or BTRFS_CLOBBERS_ALL=1 :) unconditionally, I think.
So, I think -f option is almost meaningless.

Quoted 2 lines

Otherwise, I guess I think:

Quoted 7 lines
Quoted 3 lines

WARNING! - Btrfs v0.20-rc1-212-gf6ef8b5 IS EXPERIMENTAL

and we need to expect that things might change ...

EXPERIMENTAL... It's certainly so.
However, I think that we should not add the option that it troubles a lot of people.

Well, I sent it as an RFC. Chris merged it; I'll defer to his judgement.

Thanks,
-Eric

Quoted 19 lines

Thanks,
Tsutomu

Quoted 14 lines

-Eric

Quoted 10 lines

Thanks,
Tsutomu

Quoted 6 lines

time period of 100 years where the -f option is tolerated and ignored, and then in 2113 we require that the users add the -f option?

(Just had to do this string replacement everywhere, and had to add -f to xfstest's _scratch_mkfs in common.rc as well). Sigh.

#8 lore raw
Quoted 54 lines · Eric Sandeen
Quoted 53 lines · Tsutomu Itoh
Quoted 33 lines · Eric Sandeen
Quoted 32 lines · Tsutomu Itoh
Quoted 17 lines · Stefan Behrens
Quoted 16 lines · Eric Sandeen

The core of this is shamelessly stolen from xfsprogs.

Use blkid to detect an existing filesystem or partition table on any of the target devices. If something is found, require the '-f' option to overwrite it, hopefully avoiding disaster due to mistyped devicenames, etc.

# mkfs.btrfs /dev/sda1

WARNING! - Btrfs v0.20-rc1-59-gd00279c-dirty IS EXPERIMENTAL WARNING! - see http://btrfs.wiki.kernel.org before using

/dev/sda1 appears to contain an existing filesystem (xfs). Use the -f option to force overwrite.
#

Quoted 6 lines
Quoted 3 lines

This does introduce a requirement on libblkid.

This means that it is now required to change all occurrences of "mkfs.btrfs" to "mkfs.btrfs -f" everywhere. Can't we first establish a

I also think so.
It means -f is not significant to me, I think.
(Most of my test scripts fails without -f. So I'll always type "mkfs.btrfs -f")

Therefore I want you to revert commit:2a2d8e1962e8b6cda7b0a7584f6d2fb95d442cb6.
btrfs-progs: require mkfs -f force option to overwrite filesystem or partition table

Quoted 6 lines
Quoted 1 line

How do you think about it?

What if you submit a patch to look at an environment variable, BTRFS_CLOBBERS_ALL=1 which causes it to not require -f to overwrite? Then you can just set it once at the top of your test environment, and not change every instance?

Yes. But,

Quoted 1 line
Quoted 1 line

(Most of my test scripts fails without -f. So I'll always type "mkfs.btrfs -f")

is one example.

Almost everyone types "mkfs.btrfs -f" (or BTRFS_CLOBBERS_ALL=1 :) unconditionally, I think.
So, I think -f option is almost meaningless.

Quoted 4 lines

Otherwise, I guess I think:

WARNING! - Btrfs v0.20-rc1-212-gf6ef8b5 IS EXPERIMENTAL

Quoted 7 lines
Quoted 5 lines
Quoted 1 line

and we need to expect that things might change ...

EXPERIMENTAL... It's certainly so.
However, I think that we should not add the option that it troubles a lot of people.

Well, I sent it as an RFC. Chris merged it; I'll defer to his judgement.

Agreed. So, I sent revert request to Chris :)

Thanks,
Tsutomu

Quoted 19 lines

Thanks,
-Eric

Quoted 15 lines

Thanks,
Tsutomu

Quoted 12 lines

-Eric

Quoted 9 lines

Thanks,
Tsutomu

Quoted 6 lines

time period of 100 years where the -f option is tolerated and ignored, and then in 2113 we require that the users add the -f option?

(Just had to do this string replacement everywhere, and had to add -f to xfstest's _scratch_mkfs in common.rc as well). Sigh.

#9 lore raw
Quoted 9 lines · Tsutomu Itoh
Quoted 6 lines · Eric Sandeen
Quoted 3 lines · Tsutomu Itoh

EXPERIMENTAL... It's certainly so.
However, I think that we should not add the option that it troubles a lot of people.

Well, I sent it as an RFC. Chris merged it; I'll defer to his judgement.

Agreed. So, I sent revert request to Chris :)

Where? I want to NACK the revert - this is not a matter to joke about.

You're all smart enough to know how to use shell aliases and script variables, so this "need to type -f all the time" argument holds absolutely no weight at all. Further, most of the time you're working on systems that don't hold any data you care about and so the consequences of a mistake are very minor.

However, users often make mistakes and we have to take that into account when deciding on the default behaviour of our tools. Tools that destroy data *must* err on the side of conservative default behaviour simply because of the fact that destroying the wrong data can have catastrophic consequences. It's not your data that is being destroyed, but it is data that you have a *responsibility to safeguard* as a filesystem developer.

Think about it this way: how happy would an important customer be if they decided that *you* were directly responsible for a major data loss incident because they found it would have been prevented by the "-f" patch? I don't think that the explanation of "it was inconvenient to me" would be an acceptable answer.....

Cheers,

Dave.

Signature · 2 lines
Dave Chinner
[email protected]
#10 lore raw
Quoted 7 lines · Dave Chinner
Quoted 6 lines · Tsutomu Itoh
Quoted 5 lines · Eric Sandeen
Quoted 3 lines · Tsutomu Itoh

EXPERIMENTAL... It's certainly so.
However, I think that we should not add the option that it troubles a lot of people.

Quoted 6 lines
Quoted 3 lines
Quoted 1 line

Well, I sent it as an RFC. Chris merged it; I'll defer to his judgement.

Agreed. So, I sent revert request to Chris :)

Where? I want to NACK the revert - this is not a matter to joke about.

I apologize for my childish expression.
I'll also defer to Chris's judgement.

Thanks,
Tsutomu

Quoted 25 lines

You're all smart enough to know how to use shell aliases and script
variables, so this "need to type -f all the time" argument holds
absolutely no weight at all. Further, most of the time you're
working on systems that donMy childish expression is mistaken. 't hold any data you care about and so the consequences of a mistake are very minor.

However, users often make mistakes and we have to take that into account when deciding on the default behaviour of our tools. Tools that destroy data *must* err on the side of conservative default behaviour simply because of the fact that destroying the wrong data can have catastrophic consequences. It's not your data that is being destroyed, but it is data that you have a *responsibility to safeguard* as a filesystem developer.

Think about it this way: how happy would an important customer be if they decided that *you* were directly responsible for a major data loss incident because they found it would have been prevented by the "-f" patch? I don't think that the explanation of "it was inconvenient to me" would be an acceptable answer.....

Cheers,

Dave.

#12 lore raw
In reply to Tsutomu Itoh

Am Dienstag, 26. Februar 2013 schrieb Tsutomu Itoh:

Quoted 13 lines
Quoted 6 lines
Quoted 6 lines

Therefore I want you to revert
commit:2a2d8e1962e8b6cda7b0a7584f6d2fb95d442cb6.

btrfs-progs: require mkfs -f force option to overwrite filesystem
or partition table

Quoted 6 lines
Quoted 1 line

How do you think about it?

What if you submit a patch to look at an environment variable, BTRFS_CLOBBERS_ALL=1 which causes it to not require -f to overwrite? Then you can just set it once at the top of your test environment, and not change every instance?

Quoted 7 lines

Yes. But,

Quoted 1 line
Quoted 1 line

(Most of my test scripts fails without -f. So I'll always type

"mkfs.btrfs -f") is one example.

Almost everyone types "mkfs.btrfs -f" (or BTRFS_CLOBBERS_ALL=1 :) unconditionally, I think.
So, I think -f option is almost meaningless.

No.

I don´t.

And I teach not to in my trainings as well.

Everyone who uses rm -rf by default even just for deleting a single file does it as long as he or she deleted his / her home directory or something.

Ciao,

Signature · 2 lines
Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7
#15 lore raw
In reply to Tsutomu Itoh
Quoted 9 lines · Tsutomu Itoh
Quoted 2 lines

This means that it is now required to change all occurrences of "mkfs.btrfs" to "mkfs.btrfs -f" everywhere. Can't we first establish a

I also think so.
It means -f is not significant to me, I think.
(Most of my test scripts fails without -f. So I'll always type "mkfs.btrfs -f")

Therefore I want you to revert commit:2a2d8e1962e8b6cda7b0a7584f6d2fb95d442cb6.
btrfs-progs: require mkfs -f force option to overwrite filesystem or partition table

I personally don't want to see it reverted, but also very much understand the pain to tweak all existing test scripts that rely on direct success of plain mkfs. To resolve this I'm going to use the attached script, that intsalls itself in place of mkfs.btrfs and act's as we were used to. This is intended to help developers and is not for the end user.

david

first use: mkfs.btrfs.wrapper install

---

$ cat mkfs.btrfs.wrapper
#!/bin/sh

if [ $# = 1 -a $1 = 'install' ]; then
echo "Install mode"
mkfs=type -p mkfs.btrfs
if [ $? != 0 ]; then
echo "Cannot find mkfs.btrfs"
exit 1
fi
if file $mkfs | grep -q 'ELF .* executable'; then
echo "Moving original mkfs to ${mkfs}.real" mv $mkfs ${mkfs}.real
echo "Copying myself to $mkfs"
cp $0 $mkfs
chmod 755 $mkfs
echo "Have a nice day"
exit 0
else
echo "$mkfs is not a binary, will not overwrite" exit 1
fi
fi

mkfs=type -p mkfs.btrfs.real
if [ $? != 0 ]; then
echo "Cannot find mkfs.btrfs.real, install the wrapper properly" exit 1
fi
force=

if grep -q 'Use the -f option to force overwrite' $mkfs; then

	force='-f'

fi

$mkfs $force "$@"
---

#13 lore raw
In reply to Martin Steigerwald
Quoted 18 lines · Martin Steigerwald

Am Dienstag, 26. Februar 2013 schrieb Tsutomu Itoh:

Quoted 17 lines
Quoted 6 lines
Quoted 6 lines

Therefore I want you to revert
commit:2a2d8e1962e8b6cda7b0a7584f6d2fb95d442cb6.

btrfs-progs: require mkfs -f force option to overwrite filesystem
or partition table

Quoted 6 lines
Quoted 1 line

How do you think about it?

What if you submit a patch to look at an environment variable, BTRFS_CLOBBERS_ALL=1 which causes it to not require -f to overwrite? Then you can just set it once at the top of your test environment, and not change every instance?

Yes. But,

Quoted 1 line
Quoted 1 line

(Most of my test scripts fails without -f. So I'll always type

"mkfs.btrfs -f") is one example.

Quoted 7 lines
Quoted 3 lines

Almost everyone types "mkfs.btrfs -f" (or BTRFS_CLOBBERS_ALL=1 :) unconditionally, I think.
So, I think -f option is almost meaningless.

No.

I don´t.

me too

Quoted 5 lines

And I teach not to in my trainings as well.

Everyone who uses rm -rf by default even just for deleting a single file does it as long as he or she deleted his / her home directory or something.

Unfortunately the "rm -rf" is a different case. Removing a directory is a common case. We should not be forced to use the -f for common case. A '-f' flag should be used only in "uncommon" case (like *re*format a disk or a test-suite)...

However I think that '-f' is good for mkfs.btrfs.

Quoted 2 lines

Ciao,

Signature · 2 lines
gpg @keyserver.linux.it: Goffredo Baroncelli (kreijackATinwind.it>
Key fingerprint BBF5 1610 0B64 DAC6 5F7D  17B2 0EDA 9B37 8B82 E0B5
#16 lore raw
In reply to Stefan Behrens

On 2/20/13 9:37 AM, Stefan Behrens wrote:

...

Quoted 7 lines

This means that it is now required to change all occurrences of "mkfs.btrfs" to "mkfs.btrfs -f" everywhere. Can't we first establish a time period of 100 years where the -f option is tolerated and ignored, and then in 2113 we require that the users add the -f option?

(Just had to do this string replacement everywhere, and had to add -f to xfstest's _scratch_mkfs in common.rc as well). Sigh.

I'll send a patch today to make xfstests handle both variants cleanly, I should have done that earlier, sorry.

-Eric

#14 lore raw
In reply to Goffredo Baroncelli

Am Dienstag, 26. Februar 2013 schrieb Goffredo Baroncelli:

Quoted 8 lines
Quoted 7 lines

And I teach not to in my trainings as well.

Everyone who uses rm -rf by default even just for deleting a single file does it as long as he or she deleted his / her home directory or something.

Quoted 6 lines

Unfortunately the "rm -rf" is a different case. Removing a directory is a common case. We should not be forced to use the -f for common case. A '-f' flag should be used only in "uncommon" case (like *re*format a disk or a test-suite)...

However I think that '-f' is good for mkfs.btrfs.

I don´t get you on this one:

artin@merkaba:~/Zeit> export LANG=C
martin@merkaba:~/Zeit> mkdir test
martin@merkaba:~/Zeit> rm test
rm: cannot remove 'test': Is a directory
martin@merkaba:~/Zeit#1> rm -f test
rm: cannot remove 'test': Is a directory
martin@merkaba:~/Zeit#1> rm -r test
martin@merkaba:~/Zeit> mkdir test
martin@merkaba:~/Zeit> rmdir test
martin@merkaba:~/Zeit>

So you do not need -f to remove a directory, but either rm -r or rmdir (if its empty).

I think that special casing deleting empty directories with rmdir command doesn´t make much sense. The most important distinction there IMHO is to recurse or not to recurse. Thus I would let rm also delete empty directories and remove rmdir altogether or alias it to rm (without -r!). Thats how it was done on AmigaOS delete command. Delete would delete files and empty dirs, and recurse only if ALL option was given.

rm -f doesn´t work anyway if parent directory has write permission removed:

martin@merkaba:~/Zeit> mkdir -p test/test martin@merkaba:~/Zeit> chmod a-w test martin@merkaba:~/Zeit> cd test martin@merkaba:~/Zeit> rm -r test
rm: descend into write-protected directory 'test'? y
rm: cannot remove 'test/test': Permission denied martin@merkaba:~/Zeit> rm -rf test
rm: cannot remove 'test/test': Permission denied

-f, --force
ignore nonexistent files and arguments, never prompt

Only in the case the directory itself is write-protected rm -f makes the prompt go away:

martin@merkaba:~/Zeit> mkdir test
martin@merkaba:~/Zeit> chmod a-w test
martin@merkaba:~/Zeit> rm -r test
rm: remove write-protected directory 'test'? y
martin@merkaba:~/Zeit> mkdir -m a-w test
martin@merkaba:~/Zeit> ls -ld test
dr-xr-xr-x 1 martin martin 0 Feb 26 22:12 test
martin@merkaba:~/Zeit> rm -rf test

But on skipping write protection -f is justified I think.

But more so since this is a empty directory and the parent directory has write protection, rmdir will remove it without -f (which it doesn´t support anyway):

martin@merkaba:~/Zeit> mkdir -m a-w test
martin@merkaba:~/Zeit> rmdir test
martin@merkaba:~/Zeit>

Thanks,

Signature · 2 lines
Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7
#11 lore raw
In reply to Dave Chinner

Hi,

writing from a user perspective :

The "-f" option would be highly appreciated. I agree with Dave Chinner and I guess most people who operate maschines do as well.

Cheers,
Markus

Quoted 45 lines · Dave Chinner
Quoted 9 lines · Tsutomu Itoh
Quoted 6 lines · Eric Sandeen
Quoted 3 lines · Tsutomu Itoh

EXPERIMENTAL... It's certainly so.
However, I think that we should not add the option that it troubles a lot of people.

Well, I sent it as an RFC. Chris merged it; I'll defer to his judgement.

Agreed. So, I sent revert request to Chris :)

Where? I want to NACK the revert - this is not a matter to joke about.

You're all smart enough to know how to use shell aliases and script variables, so this "need to type -f all the time" argument holds absolutely no weight at all. Further, most of the time you're working on systems that don't hold any data you care about and so the consequences of a mistake are very minor.

However, users often make mistakes and we have to take that into account when deciding on the default behaviour of our tools. Tools that destroy data *must* err on the side of conservative default behaviour simply because of the fact that destroying the wrong data can have catastrophic consequences. It's not your data that is being destroyed, but it is data that you have a *responsibility to safeguard* as a filesystem developer.

Think about it this way: how happy would an important customer be if they decided that *you* were directly responsible for a major data loss incident because they found it would have been prevented by the "-f" patch? I don't think that the explanation of "it was inconvenient to me" would be an acceptable answer.....

Cheers,

Dave.

Signature · 6 lines
Dave Chinner
[email protected]
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html