Remove -f from mkswap

[GustavoPeredo]

SUMMARY

FIXES #7946 Simply removes the -f flag from the variable MKFS_FORCE_FLAGS variable

ISSUE TYPE

• Feature Pull Request

COMPONENT NAME

filesystem

ADDITIONAL INFORMATION

[ansibullbot]

cc @abulimov @pilou- @quidame click here for bot help

[Thulium-Drake]

I think this will break the ability to force create a swap file on a system that does not run busybox. So I doubt this is a good way forward. If busybox does not implement -f to force creation, I think it needs to be addressed differently.

[GustavoPeredo]

@Thulium-Drake This is something I try to clarify in my original issue #7946 , the force option doesn't mean "format anyway" similar to other filesystem utilities force options; the use cases seem exclusively:

• Erasing the first block device
• Requesting a larger swap space than what's available on the device

It's possible to format a swapfile using mkswap /path/to/file source

I apologize in advance if your comment doesn't concern these cases but something else, if that's the case, could you give an example where it's needed to force create a swap file?

[pilou-]

When a filesystem is created by this module, regardless of the force module parameter, the force flags switches (-f for the swap partitions) are used (since always).

Should not the force flags be used only when the force module parameter is enabled?

[felixfontein]

Erasing the first block device

I think that this is the main point of passing -f. The way I understand the mkswap man page, "mkswap will refuse to erase the first block on a device with a partition table" if this option is not supplied. It still succeeds, but prints a warning like

mkswap: foo: warning: don't erase bootbits sectors
        (gpt partition table detected). Use -f to force.

(foo is the filename of a 100 MB file I created and created a partition table in with fdisk before calling mkswap foo).

It doesn't seem to care about filesystem signatures (when I ran mkfs.ext2 foo first it only said "mkswap: foo: warning: wiping old ext2 signature." when calling without -f).

Should not the force flags be used only when the force module parameter is enabled?

I think not. The way I understand the module and its force parameter is that by default, the module checks whether a filesystem already exists, and doesn't overwrite it unless you provide force. The "filesystem already exists" check of the mkfs commands is not really used, and that's why -f is always passed.

[pilou-]

Should not the force flags be used only when the force module parameter is enabled?

I think not. The way I understand the module and its force parameter is that by default, the module checks whether a filesystem already exists, and doesn't overwrite it unless you provide force. The "filesystem already exists" check of the mkfs commands is not really used, and that's why -f is always passed.

That makes sense :)

Then a fix would be to detect if Busybox is used and if so, remove the -f switch from the Swap.MKFS_FORCE_FLAGS.

[GustavoPeredo]

Removing "-f" just seems like the better option. The user can supply "-f" in the flags parameter if they wish to erase the first block, but in most cases this is really not needed/wanted. Also it spares some CPU cycles, but that's whatevs :P

If we stick to detecting busybox, is there an already established way of checking this? Or would this need to be implemented?

[felixfontein]

If we stick to detecting busybox, is there an already established way of checking this? Or would this need to be implemented?

This would need to be implemented, I think. I would add a __init__() function to Swap, similar to VFAT.__init__(), that runs mkswap --version and checks whether its stderr output has a line starting with BusyBox .

[GustavoPeredo]

Would you say that's the best approach and the way to go?

[felixfontein]

I'm not sure whether it's the best way forward, but it is one that looks implementable with a not too large amount of effort, and it looks safe in that it doesn't break backwards compatibility.